We all have our trusted indices for judging other people's drunkenness. Perhaps it's when the eyeballs start floating about as if under the control of a clumsy puppeteer. Or maybe the effusive 'you know I love you' delivered with a trickle of dribble. However, the vast majority of studies find that lay people, police officers and bartenders are in fact hopeless at distinguishing a drunk person from a sober one, at least at moderate levels of intoxication. To take just one example, after watching drunk and sober people being interviewed and negotiating a stair case, bartenders rated them as slightly, moderately or very drunk with an accuracy of just 25 per cent.
It's a similar story when participants are equipped with more structured means of detecting drunkenness. One 1958 study, for example, found no relation between doctors' assessments of people's intoxication (based on pulse rate, general appearance, gait and mental status) and the subsequent performance of those people on a driving course.
Rubenzer also looked at the evidence for specific indicators of intoxication. Alcohol causes reddening of the eyes, the literature shows, but the association between intoxication level and onset or amount of redness is unreliable. Another indicator is smell. The more a person has drunk, the more likely that their breath will be judged by observers to smell of alcohol. However, this indicator is hampered by the lack of a scientific explanation (alcohol has no odour), not to mention the risk of contamination by food smells. Speech slowing and slurring is another sign of intoxication but people are only modestly accurate at using this as a measure. Predictably enough, impaired walking, the last of the specific indicators, tends to increase the more a person has drunk but it only becomes reliable at very high intoxication levels.
The review finishes by looking at established 'sobriety tests': Nystagmus (jerky eye movements when following a moving target); the Romberg (whether a person sways or falls when they stand, eyes closed, with their feet together, arms at their sides); the Finger to Nose; the Finger to Finger; Saying the Alphabet; and the Hand Pat (alternating between clapping with the palms and backs of hands). In summary, performance on these tests does tend to decline as alcohol intake increases but the evidence for this at lower levels of intoxication is mixed and false positives (sober people categorised as drunk) are a frequent occurrence.
'...[J]udging low to moderate levels of intoxication in strangers is a difficult task,' Rubenzer concludes. 'A variety of professions that might be expected to show substantial skill assessing intoxication do not. [And] no behavioural or physical sign has emerged that is consistently related to a specific level of blood alcohol concentration level without large variation among individuals, with the possible exception of nystagmus.'
Rubenzer, S. (2010). Judging intoxication. Behavioral Sciences & the Law DOI: 10.1002/bsl.935