In response to Graham et al.'s  recent contribution entitled ‘Reducing intoxication among bar patrons: some lessons from prevention of drinking and driving’, we would like to highlight a sample of the salient methodological and practical pitfalls associated with developing a ‘valid and reliable measure of intoxication based on observable behavior’ (p. 5). For a more detailed discussion and contextualization of these limitations, we recommend that interested readers consult Brick & Erickson .
Simply stated, functional tolerance occurs when the brain adapts, and begins to compensate for, the influence of alcohol on both bodily function and behavior . That said, development of functional tolerance occurs at different rates for the various psychomotor functions, cognitive tasks and bodily functions that are impacted by alcohol [4-6]. Chronic alcohol users can reach blood alcohol concentration (BAC) levels that may incapacitate or kill others (i.e. BACs ≥ 400–500 mg/dl), while at the same time masking the accompanying visual signs of intoxication that less experienced drinkers may exhibit [7-14]. Thus, visually determining intoxication is dependent not only upon (i) the individual characteristics of the patron, but also (ii) the task being observed and (iii) the tolerance level of the person performing the tasks.
To further complicate visual identification of intoxication, it is noteworthy that tolerance can be both (i) dependent upon the setting in which the behaviour occurs and (ii) learned. Tolerance to the effects of alcohol can be accelerated if alcohol is consumed within the same environmental context, or accompanied by the same external stimuli [3, 15-18]. Additionally, if tasks are practised while drinking, it is possible that tolerance can be learned or behaviorally augmented . In other words, ‘development of ethanol tolerance is sensitive to the same training procedures which facilitate drug-free learning of a task’ (, p. 329). Tolerance associated with a specific context/environment or a certain task, however, is not transferrable to novel stimuli or unpractised conditions [20, 21].
Sensitivity and specificity
Depending upon the outcome of interest, there are varying levels of acceptable sensitivity and specificity. Unfortunately, the current evidence-base suggests that a visual test for intoxication would lack high rates of either. For instance, after watching a target enter a room, sit in a chair, engage in a standard interview and subsequently exit, four groups of social drinkers were asked to rate four different targets on level of intoxication [i.e. sober, moderately intoxicated (50 mg/dl), or very intoxicated/legally drunk (100 mg/dl)]. Only four of the 16 categorical assessments were accurate, none of which included the legally intoxicated targets . Physicians' and emergency nurses' ability to detect acute alcohol intoxication among trauma patients has also exhibited poor sensitivity (23% of patients with a BAC > 100 mg/dl were not identified) and specificity rates (patients were more likely to be falsely classified as intoxicated if young, male, disheveled, uninsured or having a low income) . While physical manifestations (i.e. staggering gait, glazed eyes, slurred speech) have demonstrated some utility in determining intoxication, these manifestations—individually and in combination—were examined at excessive intoxication levels (i.e. BAC of 0.15% and higher) . That said, even at blood alcohol concentrations above 100 mg/dl, only small proportions (21%) of drivers (n = 1115) were identified correctly as intoxicated . Thus, among average drinkers, ‘signs of visible intoxication are not readily observed at BACs that currently define intoxicated driving (80 md/dl)’ (, p. 1498). When considering chronic drinkers specifically, ‘at BACs of less than the 150 mg/dl range, most (i.e. >50%) will not appear visibly intoxicated’ (, p. 1498).
Previous attempts to establish an observational instrument
It is noteworthy that an observational instrument designed to assess alcohol intoxication visually, the 11-item Alcohol Symptom Checklist (ASC), has already been developed . Developers of the ASC contend that initial psychometric testing confirms ‘the scale's reliability, its interrater consistency, and its predictive validity’ (, p. 459). Subsequent evaluation and testing, however, documented exceedingly low correlations between the ASC ratings and a plasma alcohol concentration . More recent testing of the ASC with patients presenting at an emergency department confirmed the questionable utility of observational instruments, concluding that ‘outward physical signs of intoxication do not correlate well with BACs as measured by alcohol testing’ (, p. 4).
Overall, there is limited evidence that current premise-level interventions will reduce intoxication . Thus, as Graham et al.  assert, rethinking our approach to reducing patron intoxication levels is both necessary and recommended. Unfortunately, the ability to detect alcohol intoxication visually, even among trained professionals such as police officers, physicians and alcohol counselors, has been erratic at best ([2, 22, 23, 30-32]). Inability to reliably assess intoxication visually is due to the fact that ‘no behavioral or physical sign has emerged that is consistently related to a specific level of BAC without large variation among individuals’ (, p. 134). Given this constellation of factors, building future initiatives on the unreliable shifting sands of subjective determinations of visual intoxication does nothing to further future research efforts and may, in fact, serve as a hindrance.
Declaration of interests