Schaffer Online Library of Drug Policy Sign the Resolution
Contents | Feedback | Search
DRCNet Home
| Join DRCNet
DRCNet Library | Schaffer Library | Drugs and Driving 


Examining Variables Associated with Sampling for Breath Alcohol Analysis

C.M. Bell and H.J. Flack

Victoria Forensic Science Centre, Forensic Drive, Macleod, Victoria, 3085, Australia


The effects on BAC reading caused by differences in volume of sample delivered, exhaled breath temperature, and alteration of breathing style just prior to delivery, were studied. The variables of BAC, volume, temperature and pressure were measured over the delivery of a breath sample at a rate of 4 times per second in 14 drinking human subjects using 2 specially modified Drger Alcotest 7110 breath alcohol analysers. BAC readings at 0.5 and 1.0 L delivery volume were, on average, 804 and 932 % of the 1.5 L BAC reading respectively. Also, the BAC readings at 0.5, 1.0 and 1.5 L delivery volume were, on average, 768, 865 and 915 % of the final BAC reading respectively. A trend of lower percentage of final BAC values with increasing lung vital capacity was observed. Hyperventilation and breath holding for ca. 10 seconds just prior to sample delivery respectively decreased and increased the breath temperature, BAC and percentage of final BAC values. Standardisation of the results of breath analyses to 34.0C resulted in smoother blood alcohol decay profiles, including the tests involving hyperventilation or breath holding. The correlation between simultaneous blood and breath analyses (n=31) was, on the whole, improved by standardisation of the breath analysis result to 34.0C. The differences (breath - blood) between simultaneous blood and breath analyses were, on average, -0.00850.0070 and -0.01360.0074 g/100 mL and the calculated blood/breath partition ratio values were 2509150 and 2336172 for corrected and uncorrected data respectively. The cooling effect of the mouthpiece on the breath sample was measured as 1.0C, using both breath from a human subject and simulated breath samples.


Other studies have already reported many findings in relation to breath sampling for alcohol analysis. Dubowski and Essary (1983) and Dubowski (1974) have reported the range of exhaled breath temperatures, typical volumes of sample delivered, and pressure requirements to provide samples to particular breath alcohol analysers. Jones (1982a) stated that 70% of a subject's forced vital capacity (FVC) must be provided before a breath alcohol concentration plateau develops. Fox and Hayward (1987, 1989) investigated the effect of hypo- and hyper-thermia upon the measurement of blood alcohol concentration (BAC) by breath alcohol analysis and suggested the use of a temperature correction factor. Many workers (Mulder and Neuteboom, 1987; Fox and Pelch, 1969; Jones, 1982b) have studied the effects of hyper- and hypo-ventilation upon the result obtained by breath alcohol analysis, and Jones (1982c) has also examined the effects of temperature and humidity of inspired air upon the same.

Modern infra-red (IR) based breath alcohol analysers enable the production of alcohol concentration profiles for the term of the breath sample delivery. Gullberg (1989, 1990) has produced and mathematically analysed such profiles. The use of profile monitoring as an aid to breath sample integrity is well known (ALCOMAT Operating Instructions; Victoria Police Breath Analysis Manual, 1993). Sleymeyer (1987) has used a novel approach to improve such a monitoring system fitted in the Siemens Alcomatr instrument software.

Gullberg (1987, 1988) has also reported upon the procedure of analysing 2 separate breath samples (i.e. duplicate breath testing), discussing the significance of differences between two consecutive breathtest results. Duplicate breath testing has been adopted in the United Kingdom (Home Office and Forensic Science Service, 1993) and elsewhere and is a procedure recommended by many workers and organisations (Jones, 1990; Orhganisation Internationale, 1988).

The authors have reported (Bell, 1995a, Bell, 1995b) on a breath alcohol measurement system that provided "continuous" (actual sampling rate 4Hz) measurement of 4 variables (i.e. BAC, volume, temperature and pressure). To date, the "continuous" and simultaneous measurement of all the abovementioned variables has not been undertaken and reported in the literature, nor used in the study of issues relevant to breath alcohol analysis. One outcome from tests using this system, concerning breath temperature measurement and the effect of the mouthpiece, has already been reported (Bell, 1995b).

The "continuous" measurement system has allowed several issues and presumptions in breath alcohol analysis to be examined in new ways. Breath alcohol profiles can be generated in terms of BAC versus volume and examined in terms of minimum breath volume requirements, efficacy and possible improvement of profile monitoring, and the effects of abnormal sampling conditions (eg. mouth alcohol and hyperventilation). The influence of exhaled breath temperature can be examined in the light of recommendations to apply a temperature correction factor to breath alcohol measurements. Furthermore, blood versus breath alcohol comparisons may provide a better indication of the range of blood/breath partition ratios for alcohol, given more closely controlled sampling parameters (by way of temperature correction). Prior studies have not previously accounted for the variations in "apparent" blood /breath partition ratios for alcohol caused by variations in the exhaled breath temperature. This article is an opportunity to present some data and findings from tests conducted upon drinking human subjects, and to address these issues with the benefit provided by the "continuous" measurement facility.


Subjects of both gender were selected for drinking experiments. Only two subjects were studied at any one time. Prior to alcohol consumption, informed consent and particular subject characteristics (eg. sex, age, weight, height, state of health, smoking history and lung function) were obtained. The drinking period was limited to approximately 1.5 hours and the beverage consumed was of the subject's choice (within reason). Snacks were provided during the drinking period as well as a substantial lunch later on in the day. All subjects were tested (by breath test) to determine their pre-consumption BAC. Breath samples were provided by the subjects on a regular basis throughout the experiment, including during the initial drinking period. Disposable Products brand mouthpieces were used in all tests. Breath samples were provided to a system that carried out "continuous" measurements of BAC, volume, temperature and pressure of the sample delivered. This system is described in more detail elsewhere (Bell, 1995a), along with details on its actual performance. Two such systems were utilised in these drinking experiments.

Subjects were instructed to provide breath samples in a normal fashion, such that they inhaled slightly, then exhaled continuously and continued the exhalation until they had comfortably emptied their lungs. Excessive inspiration, as well as overstraining upon expiration (with the intent of attaining a higher total delivery volume) was discouraged. Other breathing techniques were employed in the latter stages of the experiment (i.e. when the subject's BAC was declining). These were hyperventilation or breath holding for 10 seconds just prior to sample delivery, with, as for normal deliveries, a continuous, comfortable exhalation. These techniques were performed adjacent to a normal type delivery for comparative purposes.

Blood samples were obtained in the latter stages of the experiment whilst simultaneously obtaining a breath sample. Blood analyses were performed at some time later (generally less than 3 days) by gas chromatography.


Subject Testing

Eight males and six females were studied. All subjects were of good health and returned a pre-consumption BAC (by breath test) of zero. In general, alcohol consumption was steady over the drinking period. Each breath sample was collected and stored electronically as a separate data file, which consisted of an appropriate description (i.e. name, breathing type, time, date, operator, and instrument used) and the real-time download of time, BAC, volume, temperature and pressure of the breath sample at a rate of 4 Hz. Breath sample delivery times typically were of 15 seconds duration, therefore the number of tabulated real-time readings for one test was in the vicinity of 300. In some cases, between 40 and 50 breath samples were provided by alcohol-positive subjects. This information forms a database of approximately 400 breath tests (i.e. approximately 120,000 individual readings). The data includes breath samples delivered when mouth alcohol was present, normal samples, and samples delivered after hyperventilation, breath holding and consumption of hot, non-alcoholic beverages. The database has been examined to obtain answers to specific questions posed as part of this work.

BAC and Volume

Examination of multiple BAC profiles for a subject was undertaken by normalisation of the BAC reading to remove the effect of concentration differences (i.e. real-time BAC reading / final real-time BAC reading, known as the proportion of final BAC, or PBAC). The assessment of these profiles, along with other related plots, several trends were evident. Firstly, a normal breath sample delivery produced a profile that began to plateau after 0.5 L of sample had been delivered. The values of dBAC/dV and dPBAC/dV i.e. the slope of the BAC-volume and PBAC-volume curves) were, as expected, asymtotic with the volume axis, approaching a zero value with continued breath delivery. In other words, the plateau in exhalation concentration was not flat - the BAC continued to rise slowly over the exhalation. After a delivery volume of 0.5 L, the values of dBAC/dV and dPBAC/dV were generally less than 0.1 and 2 respectively. Secondly, the decline in BAC over the breath delivery caused by mouth alcohol was obvious in breath samples provided immediately after consumption of alcohol. Five minutes later, most of the profiles closely matched those for a normal breath sample (free of mouth alcohol). Therefore, low level sample corruption by mouth alcohol may produce profiles that mimic those in the normal category and may not be detected by any type of profile monitoring software fitted in an IR breath alcohol analyser. Notably, the standard Alcotest 7110 mouth alcohol detection software (still operating in the instruments used for these experiments) detected as such many (but not all) samples in the mouth alcohol (0 min) category and none in the mouth alcohol (5 min) category. The dBAC/dV and dPBAC/dV values of 0.1 and 2 respectively only provided discrimination of the gross effects of mouth alcohol. Thirdly, the shape of the profile was slightly altered by the abnormal breathing conditions (i.e. 10 seconds of hyperventilation or breath holding just prior to sample delivery) used in this study. Hyperventilation reduced the slope, or rate of increase of the BAC, whereas breath holding reduced the time (and volume) required to attain a plateau in BAC. The threshold values for dBAC/dV and dPBAC/dV (of 0.1 and 2 respectively) did not offer any discrimination in these instances.

It is well known that the volume of breath sample provided by the subject influences the final BAC reading obtained. In Australia, all States and Territories use the Alcotest 7110 for evidential breath alcohol analyses, yet the minimum breath volume requirement varies between jurisdictions.. Data relevant to the potential differences that may arise are summarised in Table 1. For normal type breath deliveries, the BAC readings at 0.5 L and 1.0 L were, on average, 80 and 93 % of the reading at 1.5 L. The normal type breath delivery style was to exhale continuously until the subject had comfortably emptied their lungs. The final BAC reading taken at this time could be taken as the maximum value that could be expected for an evidential breath test conducted under normal conditions. The readings at 0.5, 1.0 and 1.5 L were, on average, 73, 85 and 91 % of the final BAC reading obtained after a prolonged but comfortable exhalation (which could be taken as the best possible sample of alveolar air obtainable under practical conditions).

Table 1
Selected BAC and Volume Comparisons

Tests % (0.5/1.5) % (1.0/1.5) % (0.5/final) % (1.0/final) % (1.5/final)
normal 80 93 76 86 91
n=200 (4) (2) (8) (5) (5)
hyper ventilation 80 92 76 80 88
n=40 (7) (3) (11) (7) (6)
breath holding 88 96 85 92 95
n=30 (5) (2) (8) (3) (4)
% values refer to the ratio of BAC readings within one breath test eg. BAC(0.5L)/BAC(1.5L)*100
Standard deviations are in parentheses

Hyperventilation and breath holding respectively decreased and increased the calculated relative values. Furthermore, they also respectively decreased and increased the measured exhaled breath temperature and the actual final BAC reading. Data on paired normal-hyperventilation deliveries and paired normal-breath hold deliveries are provided in Table 2. Under these test conditions, the effect of hyperventilation was more pronounced than that of breath holding in that the BAC and temperaure alterations were of greater magnitude.

Table 2
Effects of Hyperventilation and Breath Holding in Paired Tests

  BAC (g/100 mL) temp. (C) BAC (g/100 mL) temp. (C) BAC (g/100 mL) BAC (%) temp. (C)
n=51* normal hyperventilation differences
mean 0.078 34.2 0.071 33.4 -0.007 -9.4 -0.8
std dev 0.028 0.7 0.027 0.8 0.004 6.2 0.6
n=35 normal breath holding differences
mean 0.072 34.1 0.076 34.5 0.004 4.5 0.4
std dev 0.025 0.7 0.027 0.7 0.004 5.0 0.4
*number of paired tests, normal / hyperventilation
** standard deviation

The impact of BAC differences caused by differences in the volume of breath sample delivered is important for jurisdictions that use duplicate breath testing. Based upon a minimum volume requirement of 1.5 L and normal-type sample delivery, the differences in volume of sample delivered may, in extreme cases, cause differences in the BAC result as great as 20 % of the higher reading . Gullberg (1987, 1988) cites a tolerance for duplicate breath testing of 0.02 g/100 mL, or 10 % of the average result of the two tests. Setting the tolerance too tight could easily lead to failure of a subject to meet the criteria simply due to variation in delivery volume or breathing style just prior to delivery. It should also be noted that the (potential) absolute effects of hyperventilation or breath holding are the combination of the difference outlined in Table 2, plus the relative effect detailed in Table 1, if the subject provides only enough breath sample to just satisfy the minimum volume requirements.

An assesmment was made of the BAC-volume relationship with respect to lung function. A trend of decreasing percentage of final BAC values was observed with increasing lung vital capacity of the subject. The trend was evident with normal and hyperventilation type breath samples, yet less pronounced for breath holding.

BAC and Temperature

Another aspect of this study was the measurement of exhaled breath temperature. Although the temperature was measured continuously over the delivery of a sample, the consideration of breath temperature measurements has been confined to the end-expiratory temperature. Variations in end-expiratory breath temperature, from test to test, and between subjects, were observed. The modifying effect of the mouthpiece, both on measured BAC and temperature, has been previously reported (Bell, 1995b). The Disposable Products mouthpiece, the type used in these drinking experiments, lowered the measured temperature by 1.0C as compared to not using a mouthpiece. This reduction in temperature was unaffected by flowrate of the sample delivery, yet was influenced by the initial mouthpiece temperature. These drinking experiments were conducted in an air conditioned room, however ambient temperature did fluctuate by up to 10C on some days. For the purposes of making a temperature correction, the initial mouthpiece temperature has been presumed to be 20C, and therefore a constant temperature drop of 1.0C allowed for in subsequent calculations. As such, the range of exhaled breath temperatures for the 14 subjects, end-expiratory after the normal type, prolonged but comfortable exhalation, and corrected by one degree for temperature modification by the mouthpiece, was 32.9-37.0 with an average of 35.0C. This range is larger than that observed by Dubowski and Essary (1983) and further study of the temperature data for subjects showed the tendency for higher measured temperatures in the warmer ambient conditions. This questions the accuracy of the measured temperatures (under these practical conditions) as a proper reflection of the exhaled breath temperatures.

Correction of BAC results to account for the variations in exhaled breath temperature has been suggested by several workers (Fox and Hayward, 1989, 1987; Schoknecht, 1993). The temperature correction of breath test results to 34.0C is not straightforward (Bell, 1995b). It should be noted that the validity of performing such a correction needs further investigation. BAC adjustment with temperature correction to 34.0C have been performed as part of this study for trial purposes only. The formula applied to perform the correction is as follows:

BAC(34.0C) = BAC(T) * 307 / (273 + T + 1)*exp(0.06583 [34.0 - (T+1)])


BAC(34.0C) is the BAC corrected to a temperature of 34.0C,
BAC(T) is the BAC result obtained by breath analysis,
and T is the measured end-expiratory breath temperature (C).

This formula contains two parts. Firstly, when a sample enters the heated compartment of a breath alcohol analyser, it expands upon warming up to the temperature of that compartment. The degree of expansion, and therefore dilution of the alcohol concentration in the sample, differs depending upon the initial sample temperature. The correction of 307/(273 + T + 1) accounts for this effect. Secondly, the change in the partition ratio for alcohol with change in temperature of the breath sample is approximated by the exponential formula prepared by Dubowski (1979) for the air/water partition ratio for alcohol. Other studies (Fox and Hayward, 1987, 1989; Schoknecht, 1988) indicate the percentage change for alcohol in the blood / breath equilibrium is similar to that outlined in the formula Dubowski cites for the air / water system. The measured temperature was adjusted by one degree to account for the modifying effect of the mouthpiece.

Temperature correction of the breath test results produced smoother blood alcohol curves. The effects of hyperventilation and breath holding were reduced by applying the correction. Closer inspection of the tests involving hyperventilation and breath holding and their adjacent normal breath samples (i.e paired tests, as described earlier in Table 2) was performed to examine the relationship of BrAC to temperature, given a (relatively) constant BAC. Paired tests were, in general, conducted within 5 minutes of each other. The percentage difference in measured BAC and the absolute difference in temperature were compared for paired tests. Linear regression analysis indicated approximately an 8% variation in measured BAC per one degree change in measured breath temperature, in good agreement with the temperature dependence reported elsewhere (Fox and Hayward, 1987, 1989; Schoknecht, 1988).

Simultaneous blood and breath sampling and the examination of alternative breathing styles was generally undertaken in the latter stages of the experiment, once the BAC was on the decline. This design was chosen to eliminate variations caused by arterio-venous differences, which are larger in the phase of alcohol absorption (Scroggie, 1962; Mason and Dubowski, 1976).

The results of simultaneous blood and breath analyses are provided in Table 3. The results of breath analyses were, in all cases, equal to or less than the corresponding blood alcohol result. The average exhaled breath temperature for these tests, after adjustment for mouthpiece cooling of one degree, was 35.0C. The correction of the breath alcohol results therefore lowered the average breath result and increased the difference between simultaneous blood and breath analyses. The two populations of difference values (corrected and uncorrected) were significantly different (p<0.05).

Table 3
Simultaneous Blood and Breath Alcohol Results

n=31 Breath result (g/100 mL) Breath temp. (C) Blood result (g/100 mL) Breath result corrected (g/100 mL) Difference* uncorrected (g/100 mL) Difference** corrected (g/100 mL) Partition ratio*** uncorrected Partition ratio corrected
mean 0.076 34.0 0.085 0.071 -0.0085 -0.0136 2336 2509
std dev 0.029 0.8 0.032 0.027 0.0074 0.0070 172 150
* Breath - blood result
** Breath, corrected - blood result
*** Blood / breath partion ratio. Calculation based upon breath analyser calibrated on the basis of a 2100/1 partition ratio

Also included in Table 3 are calculations of the blood / breath partition ratio for alcohol. The Alcotest 7110 breath alcohol analysers used in these experiments were initially calibrated on the basis of a 2100/1 blood / breath partition ratio. Notably, the partition ratio values were, on average, 2336/1, in agreement with values reported in the more recent ratio studies and summarised by Mason and Dubowski (1976). Values for the partition ratio never fell below 2100/1, indicating that such a value for evidential breath testing (under similar circumstances) would lead to underestimation of the blood alcohol result. The partition ratio values were, on average, increased by temperature correction of the breath alcohol result and their variation was reduced. The increase in the average partition ratio was also significant (p<0.05). The comparison of corrected and uncorrected data by way of partition ratios also provided an assessment that was independent of BAC.

Although the populations of corrected and uncorrected difference values and partition ratio values were significantly different, the values for the standard deviation of each population were similar. In other words, the extent of the variation normally observed in breath alcohol analysis that is attributable to variations in exhaled breath temperature has not been reduced. The spread of results was of the same magnitude, whether the result was temperature corrected or not. Only the mean value has shifted, indicating that corrected breath results further underestimate blood alcohol analyses, and that the blood / breath partition ratio for corrected data is in excess of the accepted average for the ratio as approximately 2300/1 (Mason and Dubowski, 1976).

The current practise in Australia is to calibrate breath alcohol analysers on the basis of a 2100/1 partition ratio, and therefore the procedure underestimating the co-existing BAC is already in place. Therefore, any benefits derived from measurement of exhaled breath temperature and correction to 34.0C would be limited and be outweighed by the added expense of temperature measurement circuitry. Furthermore, as discussed earlier, the validity of temperature correction at this stage, due to sample modification by the mouthpiece, is dubious.


A novel database pertaining to breath alcohol analysis was developed and used to examine several issues relevant to this field. The BAC reading by breath analysis versus volume and versus temperature were studied. The effects of hyperventialtion and breath holding were quantified, as was the effect of differences in exhaled breath temperature. The temperature correction of breath test readings to a standardised temperature of 34.0C was found to reduce the variation in consecutive results, and make breath test results further underestimate simultaneous blood test results. Standardisation of breath teperature is, however, also dependent upon the modifying effect of the mouthpiece, which needs further investigation.


This study was undertaken as part of a grant from the National Drug Crime Prevention Fund, Canberra. The basic Alcotest 7110 instrument and the mouthpieces were kindly supplied by the Victoria Police Traffic Alcohol Section. Special thanks are given to Dr. S.J. Gutowski of the Victoria Forensic Science Centre and Mr. N. Farrell, for assistance with the preparation of the paper and data analysis.


A guide to type approval procedures for evidential breath alcohol testing intruments used for road traffic law enforcement in Great Britain, Home Office and Forensic Science Service, Aldermaston, Great Britain, (1993), p13.

ALCOMAT Operating Instructions, Siemens Ltd., p16.

Bell, C.M. and Flack, H.J., "Development of a system for real-time breath alcohol analysis", Proceedings of ICADTS - T95, Adelaide. 1995a.

Bell, C.M., "What about the humble mouthpiece? Breath sample modification and implications for breath alcohol analysis", Proceedings of ICADTS - T95, Adelaide. 1995b.

Dubowski, K.M. and Essary, N.A., "Studies on breath parameters in human subjects applicable to breath-alcohol analysis", Proceedings of the 9th International Conference on Alcohol, Drugs and Traffic Safety (ICADTS), - San Juan, Puerto Rico, U.S. Dept. of Transport, (1983), p393-398.

Dubowski, K.M., "Biological aspects of breath-alcohol analysis", Clin. Chem., (1974), v20, p294-299.

Dubowski, K.M., "Breath-alcohol simulators: Scientific basis and actual performance", J. Anal. Toxicol., (1979), v3, p177.

Fox, B.H. and Pelch, A., "Breath-holding, full time and half time exhalation in breath alcohol testing", Proceedings of the 5th ICADTS, Freiburg, (1969).

Fox, G.R. and Hayward, J.S., "Effect of hyperthermia on breath-alcohol analysis", Journal of Forensic Sciences, (1989), v34, p836-841.

Fox, G.R. and Hayward, J.S., "Effect of hypothermia on breath-alcohol analysis", Journal of Forensic Sciences, (1987), v32, p320-325.

Gullberg, R.G., "Applying a data acquisition system to the analysis of breath alcohol profiles", Journal of Forensic Sciences, (1989), v29, p397-405.

Gullberg, R.G., "Duplicate breath testing: statistical vs forensic significance of the differences", Journal of the Forensic Science Society, (1987), v27, p315-319.

Gullberg, R.G., "Duplicate breath testing; some statistical analyses", Forensic Science International, (1988), v37, p205-213.

Gullberg, R.G., "The mathematical analysis of breath alcohol profiles generated during breath exhalation", J. Anal. Toxicol. (Nov/Dec, 1990), v14, p358.

Jones, A.W., "Quantitative measurements of the alcohol concentration and the temperature of breath during a prolonged exhalation", Acta Physiol. Scand., (1982a), v114, p407-412.

Jones, A.W., "How breathing technique can influence the results of breath-alcohol analysis", Med. Sci. Law, (1982b), v22, p275-280.

Jones, A.W., "Effects of temperature and humidity of inhaled air on the concentration of ethanol in a man's exhaled breath", Clinical Science, (1982c), v63, p441-445.

Jones, A.W., "Physiological aspects of breath alcohol measurement", Alcohol, Drugs and Driving, (1990), v6, p1-25.

Mason, M.F. and Dubowski, K.M., "Breath-alcohol analysis: Uses, methods, and some forensic problems - Review and opinion", Journal of Forensic Sciences, (1976), v21, p9-41.

Mulder, J.A.G., and Neuteboom, W., "The effects of hypo- and hyper-ventilation on breath-alcohol measurements", Blutalkohol, (1987), v24, 341-347.

Schoknecht, G. and Kophamel, B., "Das temperaturproblem bei der atemalkoholanalyse", Blutalkohol, (1988), v25, p345-359.

Schoknecht, G., "The influence of temperature on breath-alcohol analysis", Alcohol, Drugs and Traffic Safety - T92, Publishers Verlag TUV Rheinland, Cologne, (1993), p392.

Scroggie, J.G., "Some aspects of recent research in breath tests for alcohol", Proceedings of the 3rd International Conference on Alcohol and Road Traffic, September, 1962. British Medical Association, London.

Second preliminary draft of an international recommendation relating to evidential breath analysers, Organisation Internationale de Metrologie Legale, France, (1988), p7.

Sleymeyer, A., "A new method for assuring alveolar-equilibrated breath alcohol samples", Proceedings of the ICADTS -T86, p1235-1241.

Victoria Police Breath Analysis Manual, 1st Edition., (1993), p56.