*689 Me. Justice White

delivered the opinion of the Court.

Plaintiff in error, David Clarence Pruitt, was tried and convicted for second degree murder growing out of an automobile accident in which two pedestrians were killed. He was sentenced to serve not more than twelve nor less than ten years in the State Penitentiary. He was also tried and convicted of leaving the scene of the accident. Sentence there was set at 11 months and 29 days in the Hamilton County Workhouse, and, in addition, a fine of $100.00 was imposed. The sentences were ordered to run consecutively.

These convictions arose out of an accident occurring on October 31,1963, at about 7:45 P.M., on Wheeler Avenue in the City of Chattanooga. The plaintiff in error’s ear and another car driven by Charley Smith were involved in a collision, in the process of which two children were killed as they walked along the side of the street. Plaintiff in error immediately left the scene of the collision and drove to his home, not far up the street. The witness, Grady Smith, a passenger in the Smith vehicle, followed Pruitt and advised him, when he got out of the car at his home, that he had struck some children.

*690 Pruitt testified that lie then went into liis house and drank half of a half pint of whiskey. About an hour later he was arrested and taken to Police Headquarters. The arresting officer and another officer who observed Pruitt at that time testified that in their opinion he was intoxicated.

At 9:20 P.M., at Police Headquarters, plaintiff in error was given a Breathalyzer test for intoxication. The test was administered by Police Officer Ingle. The results of the test showed .18 per cent alcohol in plaintiff in error’s blood. By statute, T.C.A. sec. 59-1033 (1955), a concentration of at least .15 per cent alcohol by weight creates a rebuttable presumption of intoxication.

There are essentially two assignments of error:

(1) There is no competent evidence in the record that plaintiff in error Pruitt was intoxicated or had been drinking at the time of the accident; and (2) The evidence of the results of the Breathalyzer test was improperly admitted because the officer who performed the test was not competent as an expert to either administer the test or to testify as to its results, and, in addition, he did not follow the necessary precautions.

These assignments of error concern only the conviction of second degree murder. The conviction of leaving the scene of the accident has not been challenged.

Since the evidence of the Breathalyzer test goes to the sufficiency of the evidence of drunkenness at the time of the accident, we will consider whether the results of the test were properly admitted.

Of all the police service devices now in operation for testing intoxication, the Borkenstein Breathalyzer is perhaps the most recent, having been designed and *691 developed in 1954. It operates on the principle that vapor alcohol from the lungs will oxidize in a solution of potassium dichromate and 50% sulfuric acid. The solution then loses some of its original yellow color and the color change is recorded by a photoelectric cell in comparison with an identical yellow solution which has not been exposed to the alcohol.

First, the subject blows a certain volume of alveolar (deep lung) air into a cylinder-piston chamber, the cylinder measuring exactly 52.5 cubic centimeters. This volume of air, heated at about 40 to 55 degrees Centigrade, is bubbled through the test ampule of potassium dichromate and sulfuric acid. The test ampule should be at a temperature of about 65 degrees Centigrade (approximately 150 degrees Fahrenheit). From the degree of oxidation measured by the photoelectric cell, the weight of alveolar air alcohol that causes that much color change is measured.

This alveolar air alcohol percentage is then converted by a calibrated scale to indicate the blood alcohol content. The principle behind this conversion is the fact that the same amount of alcohol, by weight, found in 52.5 ec. of alveolar air is that to be found in 1/40 cc. of the subject’s blood, a ratio of 1:2100. Also, this formula is based on the fact that in the average normal individual, the carbon dioxide (CO2) content of alveolar air is 5.5 per cent. 1

The Breathalyzer is considered a reliable device for measuring intoxication. An extensive analysis of the accuracy of the Breathalyzer is reported by investigators in an article in 8 Journal of Forensic Sciences 149 (April *692 1963), cited in Gray, Attorney’s Textbook of Medicine sec. 59.11(2) (Snpp. 1963):

These investigators state that Breathalyzer readings taken on alcohol solutions of varying concentrations very closely approximate the calculated values. The standard error associated with a reading appears to increase slightly with an increase in alcohol concentration.

The authors feel that the data presented, indicating variability in Breathalyzer readings obtained under Police Department conditions by trained police operators, are within acceptable limits and compare favorably with the data found under experimental conditions.

However, the accuracy of this device does not relieve police investigators of the requirement that a competent expert operator perform the tests. Whatever the device used, this Court has held that qualified experts must operate the machine, and they, or someone else qualified, must interpret these test results in evidence before a trial court. Fortune v. State, 197 Tenn. 691, 277 S.W.2d 381 (1954).

Requirements for an expert to qualify in testifying to the results of scientific tests are very flexible. The trial judge has discretion in admitting expert testimony, at least as to the qualifications of the expert. Fortune v. State, supra; McElroy v. State, 146 Tenn. 442, 242 S.W. 883 (1922); 7 Wigmore, Evidence sec. 1926 (3d ed. 1940). But this does not mean that this Court must accept without question the discretion of the trial judge, especially where no foundation or predicate has been laid for qualification of the expert.

*693 The State is required to show that the measuring device is scientifically acceptable and accurate for the purpose for which it is used, and that the witness who presents the test results is qualified to interpret them. It is reversible error for the State to fail to qualify a witness as an expert on the operation and results of intoxication tests with scientific devices. Fortune v. State, supra; Lopez v. State, 154 Tex.Cr.R. 227, 225 S.W.2d 852 (1949). If such testimony is admitted without proper predicate and qualification, it can be shown on cross-examination that such witness is not qualified, and it is then proper to strike his testimony. If the witness were properly qualified in advance, cross-examination would only test his credibility as an expert.

We must then determine whether the portion of the record concerning the evidence presented by the State witness Ingle showed sufficient qualifications for him to testify.

A reading of cases listed in an annotation in 77 A.L.R. 2d 971 (1961) indicates there is no settled formula for determining proper qualifications for an operator of an intoxication testing device. The cases vary from one that holds two days’ training sufficient to qualify a witness, Omohundro v. Arlington County, 194 Va. 773, 75 S.E.2d 496 (1953), and to our own decision in the Fortune case, and the case of Hill v. State, 158 Tex.Cr.R. 313, 256 S.W. 2d 93 (1953), which have held that a witness who operated the device should be able to understand the mathematical formula for converting alveolar air alcohol content to blood alcohol content.

In People v. Morgan, 236 N.Y.S.2d 1014 (Misc.1962), a witness who was “not [a] medical doctor, pathologist, biologist, hematologist, physiologist, biochemist or toxi *694 eologist, ’ ’ but who had six months ’ training as a practical nurse and “on-the-job training” under a police lab director in urine analysis was held not a competent expert to testify as to transposition or interpolation from urine alcohol content to blood alcohol content.

In State v. Gregoire, 88 R.I. 401, 148 A.2d 751 (1959), a witness who had received only three hours instruction in the operation of an “alcometer” was held not competent to testify as to results of the machine he operated even though an expert research chemist and designer of the scientific instruments stated that a man of ordinary intelligence could operate it. A trained qualified person was held especially necessary “when a new device is presented for the consideration of the courts. ’ ’ However, the result in this case was specifically rejected in the cases of State v. Roberts, 102 N.H. 414, 158 A.2d 458 (1960), and City of Wichita v. Showalter, 185 Kan. 181, 341 P.2d 1001 (1959) — at least as concerns the alcometer.

Furthermore, in State v. Bailey, 184 Kan. 704, 339 P.2d 45 (1959), it was held not to be necessary that the witness who testifies as to the results be the one who also operated the machine; however, the witness must have been properly qualified as an expert, at least to some extent, and the test must have been carried out under his supervision and control.

Robert L. Honigan, in Ms book, Chemical Tests and the Law (1957), introduces his chapter on the qualifications of persons preparing and presenting chemical test evidence with this statement:

In litigation, procedures involved in conducting chemical tests to determine alcoholic influence may be divided into five principal steps: (1) the taking of a specimen of body fluid or breath from the subject person,- (2) *695 the chemical analysis of the specimen to determine its alcoholic content, (3) the translation of this result to blood alcohol concentration, (4) the presentation of all this evidence in court from the witness stand, and (5) interpretation of the result of the test from the witness stand in the courtroom. * * *

* * * For personnel who will qualify for steps (3) and (5), the police administrator will require someone who is more than just a competent technician qualified to operate a breathalizer or drunkometer. He must be an expert in the field of chemical tests generally.

There is a practical limitation on the facilities of police administration to provide trained toxicologists or pathologists with extensive educational backgrounds. Such training is not required. We do not say that an expert in the general broad field of chemical tests is a necessity here. It is not beyond practical limits, however, to provide expert technicians with an understanding of the theoretical and operative functions of a single device for testing intoxication. He must at least be able to understand why a breath test can be translated into a certain percentage of alcohol in the blood.

To trust to the procedure of simply following instructions and getting a correct reading without any understanding of the theory that makes the reading accurate would be to approve pure hearsay evidence of intoxication. Fortune v. State, supra.

It is our conclusion, then, that even though the person who operates the testing device may be shown to know adequately the steps in handling the machine, he must further have knowledge of the reasons for such operation and the scientific principle that reflects the results as an accurate reading of blood alcohol content.

*696 With the rapid increase in automobile traffic in recent years, and consequently the proportional increase in cases involving drunk driving, police departments usually have one of these technicians with a background in alcohol testing. Their training need not be in the study of why .15 per cent alcohol in the blood is presumptive evidence of intoxication. They are relieved of this responsibility in training by T.C.A. sec. 59-1033, which provides that a presumption of intoxication arises if it be found that the defendant’s blood, as shown by chemical tests of the blood, urine, or breath of the defendant contains .15 per cent or more by weight of alcohol.

Police Officer Ingle was not shown to have been qualified to present adequately the results of the Breathalyzer test. He had about one weeks’ training from two superior police officers in how to operate the machine, but there is nothing in the record to indicate his knowledge of the theory behind the machine and the principle of conversion from alveolar air alcohol content to blood alcohol content.

The State concedes in its brief that Ingle was not shown to have the type of background that this Court requires for such technicians under the authority of Fortune v. State, supra. Nevertheless, the State insists that since Fortune did not involve a Breathalyzer, that Fortune is distinguishable from the case at bar.

The case is perhaps distinguishable in that the mechanics of operating the machine were somewhat different from those in the present case. However, the principle of conversion of lung air alcohol content to blood alcohol content is the same in both cases, and in both cases there is no proof of a knowledge of this principle or the chemical process involved on the part of the police officer.

*697 There is also undisputed proof that Officer Ingle did not conduct the test properly and thus did not take precautions against the possibility of inaccurate results. It is a recognized requirement by experts in the operation of a Breathalyzer that before beginning the test, the operator must observe the person tested for at least fifteen (15) minutes before that persons blows into the instrument. Borkenstein & Smith, “The Breathalyzer and its Application”, 2 Medicine, Science, and the Law 13 (1961); Turner, “Chemical Tests for Intoxication— Prosecution Viewpoint”, 1 Trauma, No. 3, p. 19 (Oct. 1959); State v. Baker, 56 Wash. 2d 846, 355 P.2d 806 (1960).

The purpose of this waiting period is to make sure that the person tested has no foreign matter in his mouth, that he doesn’t hiccough, vomit, belch, smoke, or take another drink — all of which could produce a false reading. Also, this period is used to keep the subject quiet, as violent physical activity may produce an abnormally high carbon dioxide content which would make the reading inaccurate. Turner, supra. If there is suspected recent drinking or foreign matter in the mouth, a duplicate sample of breath should be taken in another 15 minutes for comparison with the first test. Borkenstein & Smith, supra.

Officer Ingle tested plaintiff in error almost immediately after he was called in to administer the test. He had plaintiff in error under observation for a period of only about six minutes. We think that six minutes observation was not adequate for this defendant where the authority is unanimous — even from the developer of the device himself — that a subject must be kept under observation for at least fifteen minutes. Every precau *698 tion should have been taken to produce an accurate result, especially because of the seriousness of the crime with which plaintiff in error was charged.

For these reasons we are compelled to hold that the evidence of the Breathalyzer test results was improperly admitted and should have been stricken from the record.

We are also compelled to hold that this evidence was prejudicial to plaintiff in error. The State insists that there was other positive evidence of drunkenness at the time the test was given. Two officers testified that in their opinion plaintiff in error was intoxicated at the time of arrest. The plaintiff in error admitted he drank whiskey immediately after the accident and before the arrest. But all this is evidence of drunkenness about an hour after the accident. The witness Grady Smith who observed plaintiff in error immediately after the accident testified that he did not appear to be drunk.

The police officers testifying about the intoxication of Pruitt did not come in contact with him until an hour or more after the accident occurred.

If the Breathalyzer results are offered to show that plaintiff in error was intoxicated at the time of the accident, as well as at the time of the test, such evidence should have been the major consideration in the minds of the jurors, especially since Officer Ingle was allowed to testify as to extrapolation — that is, the rate of elimination of blood alcohol over a period of time. The evidence is also especially important in that alone it produces a presumption, by statute, of intoxication, if the reading is .15 per cent or more (T.C.A. sec. 59-1033).

The record shows that there was some contact between Pruitt’s car and a car driven by Charley Smith at or *699 near the scene where the children were killed. Prnitt failed to stop, but drove to his home a short distance away. Grady Smith followed him there shortly. On direct examination he said that when he reached Pruitt that he, Pruitt, was either getting out or had gotten out of his automobile and had “one foot inside and one out and leaning up on the car glass with the door open.”

Q. Did you look at his face good?

A. Yes sir, I stood there and looked at him.

Q. And did you smell alcohol on him?

A. No sir.

Q. How close did you get to him?

A. Well, I right up on him. I didn’t smell no alcohol.

Q. What was the expression on his face?

A. Well, he just, he just look normal. He didn’t look nothing to me.

On re-direct examination Grady Smith testified that he observed the plaintiff in error for about a minute and then he was asked:

Q. Prom your, from your observation, Grady, for a period of approximately one minute, was he or was he not under the influence of an intoxicant?

A. Prom, from me looking at him I would say he were at the time.

Upon objection to this question and answer the witness volunteered: “I didn’t smell anything. I didn’t smell anything.”

As the court overruled the objection, the witness volunteered: “Prom, from — from me looking at him I *700 would say lie — I would say lie was. Q. Drunk? A. Tes, sir.”

On re-cross-examination lie repeated lie did not smell anything, meaning alcohol.

A. No sir, I didn’t smell anything, no sir.

Q. Nothing whatsoever, did you?

A. No sir, I didn’t smell anything.

Q. No reason whatsoever to base the opinion on?

A. (Witness shook his head from side to side.)

It appears beyond question then that the evidence of alcoholic content in the blood as revealed by the Breathalyzer, although made at 9:20 P.M., more than an hour and a half after the accident occurred, was prejudicial and without such evidence the jury might well have acquitted the defendant.

Underhill, Criminal Evidence sec. 152 (5th ed. 1956), states:

* * # the difficulties faced in proving the fact of intoxication by the usual modes of testimony in motor vehicle accident cases are considerable. To distinguish between the physical reaction associated with intoxication and the similar manifestations of other conditions, such as diabetes or brain concussions frequently requires the observation and analysis of an expert. Also, the accident itself may cause a condition or reaction, like shock, which is often difficult to differentiate from intoxication.

We are convinced that it was error to admit the evidence of the Breathalyzer under the facts appearing in *701 this case. The conviction is set aside and this case is remanded for a new trial.

The conviction for leaving the scene of an accident was not appealed to this Court.

The conviction for second degree murder is reversed and remanded for further proceedings consistent with this opinion.

Reversed and remanded.

Burnett, Chief Justice, and Dyer and Chattin, Justices, concur.

Authority for this brief explanation is to be found in Borkenstein & Smith, “The Breathalyzer and its Applications”, 2 Medicine, Science, and the Law 13 (1961), and Gray, Attorney’s Textbook on Medicine sec. 59.11 (Supp.1963).