Rats were trained to discriminate nicotine (0.4 mg/kg s.c.), midazolam
(0.2 mg/kg s.c.) or the combination of these drugs from saline (n = 1
0). The rats were trained to 95% accuracy in a two-bar operant procedu
re with a tandem schedule of food reinforcement. Testing with the indi
vidual drugs in the mixture-trained group showed that nicotine (85% dr
ug-appropriate responding) was a more salient component than midazolam
(47%) in the compound stimulus. The rats were tested with benzodiazep
ine and nicotine antagonists individually and in combination (mecamyla
mine 0.2-1.6 mg/kg s.c.; flumazenil 2.5-20 mg/kg i.p.). Results for th
e mixture-trained animals shelved that flumazenil had no effect on its
own, however mecamylamine on its own produced a significant but incom
plete block in doses of 0.4-1.6 mg/kg. The greater salience of the nic
otine component of the cue would explain the block by mecamylamine but
not flumazenil. The antagonists in combination produced greater block
ade than mecamylamine on its own. The selectivity of the antagonist ac
tions on the different cue components was also demonstrated. The resul
ts suggest that in drug discrimination experiments, ''false negative''
results may be obtained with antagonists when a training drug produce
s a stimulus with more than one component.