The State of Washington appeals a decision of the Snohomish County Superior Court (Superior Court) affirming a decision of the Cascade District Court (District Court) suppressing the result of a DataMaster breath test. At issue is the method of computing a final result of two breath tests yielding different results.

Mary J. Stinson (Stinson) was arrested for driving while intoxicated. She agreed to submit to a breath test. The results of the two samples as printed out by the DataMaster were .17 and .14.1 At a pretrial hearing Stinson moved to suppress the results as not being "accurate" as defined by Washington Administrative Code 448-12-220. The WAC defines an "accurate" breath test as one which "is within plus or minus ten percent of the average of the two measurements."2 The District Court *589ruled that the average of .14 and .17 is .155, but that average should be truncated as with the actual results of the breath tests, leaving a resulting average of .15. Using this number, the reading of .17 is not within the admissible range of plus or minus 10 percent, and therefore both readings were suppressed.3 In this case, the District Court pointed out that the State indicated the DataMaster stores 3-digit results in its data base, but that it did not enter this data base into evidence. The District Court also found that the way of "averaging" as set forth in the statute was ambiguous and therefore, under the rule of lenity, the result(s) or method is required to be construed in the manner most favorable to the defendant.

The State as petitioner applied for and obtained a writ of certiorari to review the ruling. The Superior Court affirmed the decision of the District Court using the same criteria. Additionally, it specifically held that the rule of lenity applied because the method of calculating the range of admissible breath samples is ambiguous and inconsistent. The Superior Court indicated in its decision that the WAC "simply defines an 'accurate' reading and does not contemplate the various methods of calculation." The Superior Court remanded the matter to the District Court for further proceedings. Upon a motion by the State, the Superior Court stayed the proceedings in the District *590Court until 30 days after this court or the Supreme Court terminates review in the case.

WAC 448-12-220 states:

The test of a person's breath for alcohol concentration by infrared test method shall consist of the person insufflating deep lung air samples at least twice into the instrument sufficient to allow two separate measurements. There will be sufficient time between the provision of each sample by the person to permit the instrument to measure each sample individually. The two breath samples supplied by the individual shall constitute one test. An accurate test will be presumed if the results of each measurement is within plus or minus ten percent of the average of the two measurements.

(Italics ours.) The word "average" is not defined in the WAC, or by statute. " 'Absent a statutory definition, words of a statute must be accorded their ordinary meaning.' " State v. Halsen, 111 Wn.2d 121, 123, 757 P.2d 531 (1988) (quoting Davis v. Department of Empl. Sec., 108 Wn.2d 272, 277, 737 P.2d 1262 (1987)).

Average is defined as "[a] mean proportion, medial sum or quantity, made out of unequal sums or quantities. Brisendine v. Skousen Bros., 48 Ariz. 416, 62 P.2d 326, 329 [(1936)]. In ordinary usage the term signifies the mean between two or more quantities, measures or numbers." Black's Law Dictionary 135 (6th ed. 1990).

In the case before the court, Stinson's two breath samples were .14 and .17. Totaled together their sum is .31. Dividing that total by 2 gives an average of .155. The average figure, .155, contains three digits instead of two. This will always be the case when the sum total of the two breath readings is odd, because it will always be divided by the even number 2.

As the State argues, the term "average" or "mean" as a general matter of mathematical precision is usually displayed to more digits than the individual figures or their sum. The State quotes from S. Meyer, Data Analysis for Scientists and Engineers 26 (1975), "the means [are] calculated to one significant figure more than the individual measurements. . . . The . . . justification, of course, is that *591the average is more precise than any of the individual values and requires more significant figures." Therefore, the State argues the truncation of the average of the two samples from .155 to .15 is incorrect for two reasons: (1) it violates the principle or justification set forth above; and (2) it is incorrect by definition.4 5We agree. Further, the action of the District Court as affirmed by the Superior Court has the effect of narrowing the range of acceptable readings, especially affecting those readings below the .10 mark. Thus, the result of the lower court's holding may prejudice those who register less than a .10 and who desire to get the result admitted in evidence.

Stinson claims the method of calculating the range of admissible breath samples is ambiguous and inconsistent. Because the DataMaster stores three digits, but only prints to the second digit, there is an implication that the third digit is insignificant. Thus, if the third digit of the average is to be used in calculating the average range of the samples, the defense argues and the courts below agreed, then the third digit of the samples should be used.

Stinson is incorrect. First, as argued by the State, it is a rule of mathematics that the average or mean may require a further significant digit. Second, the third digit of the DataMaster reading is truncated or dropped for two reasons: (1) it is to the benefit of the defendant to have the third figure truncated;5 and (2) a determination has been made that a reading to the thousandths of a gram is not especially scientifically significant.6

*592Stinson argues, as the lower courts held, that because there are two ways of calculating an average, which is not really the case, the WAC is ambiguous, and thus the rule of lenity must be applied. Even assuming arguendo that there are two ways of calculating an average, the rule of lenity does not apply in this case.

The rule of lenity requires that a criminal statute which is susceptible of more than one reasonable interpretation be construed strictly against the State and that the interpretation most favorable to the accused be adopted. State v. Gore, 101 Wn.2d 481, 486, 681 P.2d 227, 39 A.L.R.4th 975 (1984) (citing State v. Sass, 94 Wn.2d 721, 620 P.2d 79 (1980)). However, it is punitive statutes that must " 'be literally and strictly construed in favor of the accused'." Halsen, 111 Wn.2d at 123 (quoting State v. Hornaday, 105 Wn.2d 120, 127, 713 P.2d 71 (1986)). Here the WAC is not a penal statute, but an evidentiary one. There are not two different penalties imposed by WAC 448 -12-220; it is by nature a definition of the breath test and a method of determining the accuracy of a test. It sets forth the requirements for the admission of a particular type of evidence. It does not set forth a definition of a crime or a penalty.

Therefore, both lower courts were incorrect in applying the rule of lenity. This, in conjunction with the erroneous determination by each lower court that the use of the word "average" is ambiguous, compels us to reverse the decision of the lower courts as to the suppression of the results of the DataMaster breath test. We remand the case to the District Court for trial.

Scholfield and Pekelis, JJ., concur.