Police officers frequently use the presence or absence of an alcohol breath
odor for decisions on proceeding further into sobriety testing. Epidemiolo
gical studies report many false negative errors. The current study employed
20 experienced officers as observers to detect an alcohol odor from 14 sub
jects who were at blood alcohol concentrations (BACs) ranging from zero to
0.130 g/dl. Over a 4 h period, each officer had 24 opportunities to place h
is nose at the terminal end of a 6 in, tube through which subjects blew. Su
bjects were hidden behind screens with a slit for the tube to prevent any b
ut odor cues. Under these optimum conditions, odor was detected only two-th
irds of the time for BACs below 0.08 and 85% of the time for BACs at or abo
ve 0.08%. After food consumption, correct detections declined further. Offi
cers were unable to recognize whether the alcohol beverage was beer, wine,
bourbon or vodka. Odor strength estimates were unrelated to BAC levels. Est
imates of BAC level failed to rise above random guesses. These results demo
nstrate that even under optimum laboratory conditions, breath odor detectio
n is unreliable, which may account for the low detection rate found in road
side realistic conditions. (C) 1999 Elsevier Science Ltd. All rights reserv
ed.