Snap Shot
- The question: When no one measured a person’s blood alcohol concentration (BAC) at the time that matters, how do experts estimate it from a later test or from a description of the drinking, and how much confidence do those estimates deserve?
- What the research shows: The Widmark formula estimates BAC from the amount of alcohol consumed, body weight, a distribution factor, and an elimination rate; retrograde extrapolation works backward from a measured BAC by adding back the alcohol presumed eliminated in the interval. Both rest on assumptions (that absorption was complete, that elimination was constant, that the drinking history is known) and on population values that vary widely between people. One formal error analysis estimated the uncertainty of a Widmark BAC at about 21% (one standard deviation) for a worked example and recommended reporting a range of about twice that (Gullberg, 2007); another showed that the size of the range depends on the inputs in each case and must be calculated for that case (Searle, 2015). Estimates built on self-reported drinks can be wrong in either direction, and the net direction depends on the case.
- What it does not show: A calculation cannot establish a particular person’s BAC with precision, cannot show what the person remembered or how he or she appeared, and cannot establish legal incapacity. An estimate is only as good as its inputs, and a single number without a range overstates what the method can deliver.
- The law: Many courts have treated retrograde extrapolation as a generally accepted method when its assumptions are stated, and its application is tested under Rule 702 (as amended in 2023), Daubert, Kumho, Mil. R. Evid. 702 and Houser, and Ariz. R. Evid. 702; in Arizona, State ex rel. Montgomery v. Miller (App. 2014) vacated an order excluding an extrapolation that used conservative rates and a range, treating disputes about its assumptions largely as matters of weight. Arizona’s per se offense looks to the concentration within two hours of driving (A.R.S. § 28-1381(A)(2)), and Missouri v. McNeely (2013) held that the natural dissipation of alcohol does not by itself excuse a warrant for a blood draw.
- Issues for each side: Either side can legitimately offer or test an extrapolation; both should insist on stated assumptions and ranges, recognize that uncertain drink counts and an unknown absorption state can move the estimate in either direction, and avoid presenting a calculated value as a measurement.
Many cases turn on a person’s level of intoxication at a moment when no one measured it. A breath or blood sample may be taken hours after a crash, an arrest, or an alleged assault, or not at all; sometimes the only information is a witness’s account of what each person drank. Experts fill the gap with calculations. Working backward from a later test is called retrograde extrapolation (also back-calculation or relation back). Working forward from the drinks consumed uses the equation that the Swedish physiologist Erik Widmark developed in the early twentieth century.
These calculations are routine in forensic toxicology, and they can be informative. They are also among the most frequently disputed forms of alcohol testimony, because each depends on assumptions that may not hold for the person in question and on average values that vary widely. This paper explains how the calculations work, what they assume, how large the uncertainty is, and how errors can run in either direction. It replaces an older page that reproduced an online Widmark calculator; the formula is explained here in words, with the research on its limits. The basic pharmacology (absorption, distribution, and elimination) is covered in Alcohol Toxicology for Attorneys: Absorption, Distribution, Elimination, and Stages of Intoxication.
The Widmark Formula
The Widmark equation rests on a simple idea: once alcohol has been absorbed, it is spread through the body’s water, so the concentration in the blood equals the amount of alcohol divided by the effective volume into which it is diluted, minus whatever has been eliminated since drinking began. In words, the estimated BAC equals the grams of alcohol consumed, divided by the person’s body weight multiplied by a distribution factor (called r or rho), minus the elimination rate multiplied by the time since drinking began (Jones, 2019). Run in reverse, the same equation estimates how much alcohol must have been consumed to produce a measured BAC.
Each input carries uncertainty. The amount consumed depends on how many drinks there were, how large they were, and how strong; the paper Self-Reports of Drinking: Underestimation and Counting Drinks describes how often each is misjudged. Body weight is usually known. The distribution factor varies with sex, age, and body composition: Widmark’s own subjects averaged 0.68 for men and 0.55 for women, with individual values ranging from about 0.55 to 0.86 in men and 0.47 to 0.65 in women (Jones, 2008), and current reviews give typical values near 0.70 and 0.60 (Jones, 2019). Equations that estimate the factor from height, weight, age, and sex improve accuracy on average, but when compared with values measured in 173 men and 63 women, even the best left 95% ranges of about plus or minus 0.13 L/kg for men and 0.15 L/kg for women (Maskell et al., 2019). The elimination rate, discussed below, also varies several-fold.
Retrograde Extrapolation From a Later Test
Retrograde extrapolation begins with a measured BAC and adds back the alcohol presumed to have been eliminated between the event and the test. If a test two hours after an event shows 0.10 g/dL and the person eliminated alcohol at 0.015 g/dL per hour, the calculation gives 0.13 g/dL at the time of the event; at 0.020 per hour it gives 0.14. A widely cited methods paper argued that the approach is reliable when its assumptions are reasonable, recommended an elimination range of about 0.015 to 0.020 g/dL per hour, and noted that a wider range (0.010 to 0.030) accounts for unusual individuals but that estimates for a time before the peak are much less certain (Montgomery & Reasor, 1992). Jones (2010), in an evidence-based review, put the physiological range at about 0.010 to 0.035 g/dL per hour, with about 0.015 typical of moderate drinkers and about 0.019 of apprehended drivers, and in a leading forensic reference he recommended calculating with a range of rates rather than a single average (Jones, 2008).
In 2024 the Academy Standards Board published ANSI/ASB Best Practice Recommendation 122 (1st ed.) on alcohol calculations in forensic toxicology. It is a best practice recommendation, not a statute or court rule, and it covers retrograde extrapolation, forward estimation, and minimum drinks consumed; it states that expert opinions based on the calculations are outside its scope. Among its provisions, calculations are to use a range of elimination rates with a minimum range of 0.010 to 0.025 g/dL per hour, which is wider than the 0.015 to 0.020 range in Montgomery and Reasor (1992) and narrower than the physiological range in Jones (2010), and a rate derived from the person’s own two or more test results does not replace that range. A person may reasonably be assumed to be post-absorptive when the incident occurred more than two hours after drinking stopped, but not when the drinking history is unknown. Retrograde extrapolation is not to be performed on concentrations below 0.020 g/dL or on urine results, and the volume of distribution is also applied as a range (0.45 to 0.81 L/kg regardless of sex, or 0.58 to 0.83 for males and 0.43 to 0.73 for females) (Academy Standards Board, 2024). Practice at very low concentrations is not uniform, as noted above, and practitioners should confirm the current edition.
The method makes three key assumptions. First, absorption must have been complete at the time of the event, so that BAC was already falling; if the person was still absorbing alcohol, the BAC at the event may have been lower than the later test, and adding back elimination would overstate it. Peak BAC is usually reached within about an hour after the end of drinking, but some people take two hours or more, and food prolongs absorption (Jones, 2008; Jones & Jönsson, 1994). Second, elimination must have proceeded at a roughly constant rate across the interval, which generally holds above very low concentrations (Jones, 2010). Third, it assumes no alcohol was consumed between the event and the test or, if some was, that the amount is known and can be accounted for.
When the drinking history is unknown, these assumptions are harder to defend. The same forensic reference describes back-calculation as a contentious area and notes that, without information about when and how much a person drank, the BAC at the time of an event may have been rising, at a plateau, or falling (Jones, 2008). The practical consequence is that extrapolations over short intervals soon after drinking are the least certain, and extrapolations from a test taken well after drinking ended, with a known time of last drink and a range of rates, are the most defensible. Practice also varies at the low end of the curve: German forensic medicine has traditionally declined to back-calculate from very low concentrations, although a 2026 re-analysis of 87 drinking-trial subjects found that elimination remained essentially linear at those levels and proposed a modified rule (Mihan et al., 2026).
Case Example. In State ex rel. Montgomery v. Miller (App. 2014), a driver was stopped at about 2:20 a.m., and blood drawn at about 6:15 a.m. showed a BAC of 0.127. The State’s expert extrapolated back to the period within two hours of driving, assuming no food or alcohol after the stop and absorption complete within two hours of the last drink, and reported a range (0.127 to 0.177) using conservative rates. A defense expert maintained that a valid extrapolation required the driver’s eating and drinking history. The trial court excluded the testimony under Rules 702 and 403; the Arizona Court of Appeals vacated that order, holding that the method was generally accepted, that conservative rates and a range addressed the risk of overestimation, and that disputes over the absorption assumption and the missing history went mainly to weight, to be tested by cross-examination and contrary evidence. The decision illustrates how courts treat these calculations; it does not establish that any particular extrapolation is accurate.
How Large Is the Uncertainty?
Several analyses have quantified the error. Applying formal error propagation to the Widmark equation, Gullberg (2007) found that an estimated BAC of 0.120 g/dL carried an uncertainty (one standard deviation) of about 0.026, a coefficient of variation of about 21%, and recommended reporting a range of roughly plus or minus 42% (two coefficients of variation) around a BAC estimate. Searle (2015) agreed that every calculated BAC should be reported with an error range but showed that the size of that range is not a fixed percentage and must be calculated for each case from its own inputs. In Gullberg’s worked example, a calculated 0.120 g/dL with a range of about plus or minus 42% spans roughly 0.07 to 0.17 g/dL, so a calculated value is better understood as a range than as a point; the range for any other case depends on that case’s inputs.
Field data point the same way. When estimated BACs were calculated from the drinks reported by 109 people recruited near bars and compared with breath tests, the equations agreed closely with one another but explained only about 55% of the variance in measured breath alcohol, overestimated on average, and fit less well as intoxication increased (Hustad & Carey, 2005). In a laboratory drinking-game study with 40 participants, formula-based estimates tended to exceed breath readings, and the discrepancy was larger for women (Silvestri et al., 2013). A published commentary on the blackout literature (a commentary, not a study) argued that the Widmark calculation can substantially overestimate BAC after prolonged heavy drinking (Wigmore, 2014). Because underreporting of drinks, which is common at high consumption, pushes estimates the other way, the net direction of error depends on the case.
Error in Either Direction
Errors in these calculations do not all point the same way, and the direction depends on which assumption fails. An estimate tends to be too high when the person was still absorbing alcohol at the time of interest, when the distribution factor used is smaller than the person’s true value, when the elimination rate used is faster than the person’s true rate in a back-calculation from a test, or when reported drinks were overcounted or assumed to be stronger than they were. An estimate tends to be too low when drinks were undercounted or poured larger than a standard drink, when the elimination rate used is slower than the person’s true rate in a back-calculation (as the conservative choices recommended for forensic use deliberately do), or when the distribution factor used is larger than the person’s true value. Underreporting of drinks is common, especially at high consumption (Northcote & Livingston, 2011), and commercially poured wine and mixed drinks often contain more alcohol than a standard drink (Kerr et al., 2008).
Because the sources of error differ, a careful analysis states each assumption, uses a range for each uncertain input, and reports the result as a range. Conservative choices protect against overestimation in a back-calculation from a known test, but they do not protect against an unrecognized absorption phase, which can make even a conservative back-calculation too high, or against undercounted drinks in a forward calculation, which can make an estimate too low. Sex enters the calculation through the distribution factor: on average, women reach higher BACs than men from the same dose, mainly because of differences in body water (Mumenthaler et al., 1999), while elimination rates are similar or slightly faster in women (Jones, 2010). Using a man’s factor for a woman, or the reverse, introduces a predictable error.
Limits of the Evidence
The data behind elimination rates and distribution factors come mainly from controlled drinking studies of healthy volunteers, often men and often fasting, and from large samples of Scandinavian drivers who gave two samples; both differ from the long, mixed, and poorly documented drinking episodes that reach court. Women are represented in fewer subjects, although the main sex differences are consistent (Maskell et al., 2019; Jones, 2010). Error analyses model uncertainty in the inputs but cannot correct a wrong assumption about absorption or a wrong drinking history. Field validation studies of drink-based estimates used young adults near bars at moderate levels, not the very high levels at issue in many cases. Above all, these methods estimate a concentration; they do not measure memory, appearance, judgment, or capacity, and the research cannot determine a particular person’s BAC at a particular moment beyond the range its inputs support.
Questions to Ask About This Research
When an expert gives a single back-calculated value, counsel can ask what range results from the documented span of elimination rates (Jones, 2010) and why a particular rate was chosen. When the calculation assumes the person was post-absorptive, counsel can ask what is known about the time of the last drink and food, and how the estimate would change if the person was still absorbing alcohol (Jones, 2008). When a Widmark estimate relies on drink counts, counsel can ask who supplied them, how drink sizes were determined, and how the result would change with plausible undercounting or overcounting (Northcote & Livingston, 2011; Kerr et al., 2008). When a distribution factor of 0.68 or 0.55 is used, counsel can ask whether it was adjusted for the person’s build and what uncertainty remains (Maskell et al., 2019). And when an expert reports an estimate without an error range, counsel can ask what range the expert would report under the methods described by Gullberg (2007) and Searle (2015). These questions are explained in “Reading Research Claims: A Guide for Attorneys” on this site.
Retrograde Extrapolation and the Law
Federal and Military Standards
Extrapolation testimony is expert testimony governed by Federal Rule of Evidence 702, as amended December 1, 2023, which requires the proponent to show that it is more likely than not that the testimony rests on sufficient facts or data, reflects reliable principles and methods, and reflects a reliable application of those methods to the case. Daubert v. Merrell Dow Pharmaceuticals, Inc. (1993) makes the court a gatekeeper, and Kumho Tire Co. v. Carmichael (1999) extends that obligation to all expert testimony. The 2023 amendment’s emphasis on reliable application fits extrapolation disputes, which usually concern whether the facts about drinking, food, and timing are sufficient and whether the assumptions fit the person rather than whether the method is valid in general. Military Rule of Evidence 702 tracks the federal rule, and the 2023 amendment took effect in the military rule in 2025 (Joint Service Committee on Military Justice, 2025); practitioners should confirm current text. Military judges also apply the factors of United States v. Houser (1993).
Article 113, UCMJ, prohibits operating or physically controlling a vehicle, aircraft, or vessel while drunk or impaired, or when the alcohol concentration in the person’s blood or breath equals or exceeds the applicable limit (in the United States, the lesser of the state limit and 0.08 grams per 100 milliliters of blood or per 210 liters of breath). The article does not contain a within-two-hours provision like Arizona’s, so when a test is taken well after the operation, how the result relates to the time of operation is a matter of proof. Timing also affects how samples are obtained. In Missouri v. McNeely (2013), the Supreme Court held that the natural dissipation of alcohol in the blood does not create a per se exigency justifying a warrantless blood draw in every drunk-driving case; exigency is judged on the totality of the circumstances. In the military, nonconsensual extraction of body fluids without a warrant or search authorization is admissible only under the conditions of Military Rule of Evidence 312(d), including probable cause and a risk that the delay needed to obtain authorization could result in destruction of the evidence.
Arizona
Arizona Rule of Evidence 702 adopted the federal formulation in 2012 and was amended effective January 1, 2024, to track the 2023 federal amendment (Ariz. Sup. Ct. Order R-23-0004). Arizona’s per se offense applies when a person has an alcohol concentration of 0.08 or more within two hours of driving or being in actual physical control, and the concentration results from alcohol consumed before or while driving (A.R.S. § 28-1381(A)(2)); the impairment offense requires impairment to the slightest degree (A.R.S. § 28-1381(A)(1)). The two-hour provision means that extrapolation in Arizona driving cases often runs from a later test back to the two-hour window rather than to the moment of driving, as in Miller, where the court vacated the order excluding the State’s extrapolation under Rule 702 and treated disputes about absorption and the missing drinking history mainly as matters of weight. Experts may explain general principles but may not opine on whether a particular witness is truthful (State v. Lindsey, 1986, 149 Ariz. at 475), which bears on testimony that relies on, or comments on, a witness’s account of the drinking.
Special Issues in Legal Practice
The principles described above apply with equal force to every party in a legal proceeding. The science operates the same way regardless of who is asking the questions, and the research summarized here does not by itself favor either side. What differs between the parties is not the science but the vantage point — the particular places where these issues tend to surface, and the particular errors each party is best positioned to guard against. The two summaries below identify those recurring issues. They are offered for orientation, not as litigation strategy; in every case the controlling question is what the evidence in that case will, and will not, support, and the expert’s role is to inform the trier of fact rather than to advocate for either side.
Special Issues for Prosecutors
The research supports several propositions prosecutors can legitimately present. Many courts, including the Arizona Court of Appeals in Miller, have treated retrograde extrapolation as generally accepted when its assumptions are stated, and an estimate built on a known test, a documented time of last drink, conservative elimination rates, and a reported range is the kind of testimony Miller declined to exclude. Because conservative rates deliberately understate elimination, a range built on them tends, in a back-calculation from a later test, to protect against overestimation; it does not protect against an unrecognized absorption phase. Underreporting of drinks is common at high consumption, and poured drinks often exceed standard sizes, so a forward calculation from an account of the drinking may understate a person’s BAC; the same point applies to a complainant whose own account of drinking may be incomplete.
The research also counsels caution. A back-calculation assumes absorption was complete, and when drinking may have continued shortly before the event, the estimate can be too high even with conservative rates. A single value, or a range that omits uncertainty in the distribution factor and the drinking history, overstates precision, and published error analyses suggest ranges wide enough to matter at legal thresholds (Gullberg, 2007). A calculated BAC for a complainant is not evidence of incapacity by itself, and the uncertainty that applies to an accused’s estimate applies equally to a complainant’s.
Special Issues for Defense
The research supports several propositions defense counsel can legitimately present. Each input to a Widmark or back-calculation is uncertain, the distribution factor and elimination rate vary widely between people, and the method cannot establish whether a person was still absorbing alcohol without reliable information about when he or she drank and ate. Formal error analyses support reporting a range with every calculated BAC, one estimating it at about 21% (one standard deviation) for a worked example, with a recommended range of about twice that (Gullberg, 2007), and another showing that it depends on the inputs and must be calculated case by case (Searle, 2015). Drink-based estimates have overestimated measured breath alcohol on average in field validation (Hustad & Carey, 2005), and a published commentary (not a study) has argued that the Widmark approach overestimates after prolonged heavy drinking (Wigmore, 2014), while under-reporting of drinks (Northcote & Livingston, 2011) pushes estimates the other way, so the net direction depends on the case. These are legitimate subjects for examination of any expert who offers a calculation.
The research also counsels caution about the limits that apply to the defense. The uncertainty runs in both directions: undercounted drinks and larger-than-standard pours can make an estimate too low, and conservative elimination rates already tilt a back-calculation downward. Miller indicates that disputes about absorption and drinking history will often go to weight rather than admissibility in Arizona. A defense estimate offered to show that an accused was too intoxicated to form a mental state, or that a complainant was only mildly intoxicated, is subject to the same limits, and an alternative calculation without stated assumptions and a range is open to the same objections.
This paper is for educational purposes only and is not legal advice. The author is a psychologist, not an attorney; summaries of law are general and must be verified against current authority or with a licensed attorney.
References
Academy Standards Board. (2024). Best practice recommendation for performing alcohol calculations in forensic toxicology (ANSI/ASB Best Practice Recommendation 122, 1st ed.). American Academy of Forensic Sciences. https://www.aafs.org/asb-standard/best-practice-recommendation-performing-alcohol-calculations-forensic-toxicology
Gullberg, R. G. (2007). Estimating the uncertainty associated with Widmark’s equation as commonly applied in forensic toxicology. Forensic Science International, 172(1), 33–39. https://doi.org/10.1016/j.forsciint.2006.11.010
Hustad, J. T. P., & Carey, K. B. (2005). Using calculations to estimate blood alcohol concentrations for naturally occurring drinking episodes: A validity study. Journal of Studies on Alcohol, 66(1), 130–138. https://doi.org/10.15288/jsa.2005.66.130
Joint Service Committee on Military Justice. (2025). Amendments to the Military Rules of Evidence since 1984 (G. E. Maggs, Ed.; 2025 ed.). https://jsc.defense.gov/Portals/99/MRE%20Amendments%202025.pdf
Jones, A. W. (2008). Biochemical and physiological research on the disposition and fate of ethanol in the body. In J. C. Garriott (Ed.), Garriott’s medicolegal aspects of alcohol (5th ed., pp. 47–156). Lawyers & Judges Publishing.
Jones, A. W. (2010). Evidence-based survey of the elimination rates of ethanol from blood with applications in forensic casework. Forensic Science International, 200(1–3), 1–20. https://doi.org/10.1016/j.forsciint.2010.02.021
Jones, A. W. (2019). Alcohol, its absorption, distribution, metabolism, and excretion in the body and pharmacokinetic calculations. WIREs Forensic Science, 1(5), Article e1340. https://doi.org/10.1002/wfs2.1340
Jones, A. W., & Jönsson, K. A. (1994). Food-induced lowering of blood-ethanol profiles and increased rate of elimination immediately after a meal. Journal of Forensic Sciences, 39(4), 1084–1093. https://doi.org/10.1520/JFS13687J
Kerr, W. C., Patterson, D., Koenen, M. A., & Greenfield, T. K. (2008). Alcohol content variation of bar and restaurant drinks in Northern California. Alcoholism: Clinical and Experimental Research, 32(9), 1623–1629. https://doi.org/10.1111/j.1530-0277.2008.00741.x
Maskell, P. D., Jones, A. W., Savage, A., & Scott-Ham, M. (2019). Evidence based survey of the distribution volume of ethanol: Comparison of empirically determined values with anthropometric measures. Forensic Science International, 294, 124–131. https://doi.org/10.1016/j.forsciint.2018.10.033
Mihan, A. T. R., Toennes, S., Paulke, A., & Verhoff, M. A. (2026). On the possibility of retrograde calculation of blood alcohol concentrations below 0.15‰. International Journal of Legal Medicine, 140(5), 2833–2844. https://doi.org/10.1007/s00414-026-03838-6
Montgomery, M. R., & Reasor, M. J. (1992). Retrograde extrapolation of blood alcohol data: An applied approach. Journal of Toxicology and Environmental Health, 36(4), 281–292. https://doi.org/10.1080/15287399209531639
Mumenthaler, M. S., Taylor, J. L., O’Hara, R., & Yesavage, J. A. (1999). Gender differences in moderate drinking effects. Alcohol Research & Health, 23(1), 55–64.
Northcote, J., & Livingston, M. (2011). Accuracy of self-reported drinking: Observational verification of ‘last occasion’ drink estimates of young adults. Alcohol and Alcoholism, 46(6), 709–713. https://doi.org/10.1093/alcalc/agr138
Searle, J. (2015). Alcohol calculations and their uncertainty. Medicine, Science and the Law, 55(1), 58–64. https://doi.org/10.1177/0025802414524385
Silvestri, M. M., Cameron, J. M., Borsari, B., & Correia, C. J. (2013). Examining alcohol and alcohol-free versions of a Simulated Drinking Game Procedure. Journal of Studies on Alcohol and Drugs, 74(2), 329–336. https://doi.org/10.15288/jsad.2013.74.329
Wigmore, J. G. (2014). Commentary on: Pressman MR, Caudill DS. Alcohol-induced blackout as a criminal defense or mitigating factor: An evidence-based review and admissibility as scientific evidence. J Forensic Sci 2013;58(4):932–40. Journal of Forensic Sciences, 59(1), 293. https://doi.org/10.1111/1556-4029.12364
Legal Authorities
Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993).
Kumho Tire Co. v. Carmichael, 526 U.S. 137 (1999).
Missouri v. McNeely, 569 U.S. 141 (2013).
State ex rel. Montgomery v. Miller, 234 Ariz. 289, 321 P.3d 454 (App. 2014).
State v. Lindsey, 149 Ariz. 472, 720 P.2d 73 (1986).
United States v. Houser, 36 M.J. 392 (C.M.A. 1993).
10 U.S.C. § 913 (Article 113, Uniform Code of Military Justice).
Ariz. Rev. Stat. Ann. § 28-1381(A)(1)–(2).
Fed. R. Evid. 702.
Mil. R. Evid. 312(d), 702, 1102.
Ariz. R. Evid. 702.
Ariz. Sup. Ct. Order R-23-0004 (amending Ariz. R. Evid. 702, effective Jan. 1, 2024).
Related topic papers: Alcohol Toxicology for Attorneys: Absorption, Distribution, Elimination, and Stages of Intoxication · Breath Versus Blood Testing: Partition Ratios, Serum Results, and Hospital Records · Self-Reports of Drinking: Underestimation and Counting Drinks · Reconstructing a Blackout Period from Digital and Documentary Evidence · Blackouts and Blood Alcohol Concentration
Related topics from Arizona Forensic Services: Counterintuitive Testimony: Drug-Facilitated Sexual Assault: Toxicology, Detection Limits, and What the Results Can Show · Courts-Martial