Background. Previous research has been unable to show unequivocally wh
ether flumazenil can reverse completely, partially, or not at all the
memory effects of benzodiazepines. The effects of midazolam on implici
t memory are also unknown. The behavioral effects of flumazenil by its
elf, and the acute reversal of benzodiazepine effects, are also contro
versial. The current study was designed to investigate these questions
. Methods. Using a prospective randomized, double-blind crossover stud
y design, memory was assessed using both direct (free recall and recog
nition) and indirect (word completion) measures. Other cognitive effec
ts were assessed using the digit symbol substitution task. Sedation an
d other mood effects were assessed using subjective rating scales. Sev
enty-two healthy subjects were assigned to three equal groups accordin
g to the dose of midazolam received (0, 0.05, and 0.1 mg/kg). Each sub
ject received varying doses of flumazenil (0, 1, and 3 mg) in three se
ssions, at least 1 week apart. After baseline administration of the ta
sks, midazolam was administered. The assessments were repeated 20 min
later, followed by administration of flumazenil. The assessments were
repeated 5 and 30 min after administration of flumazenil. After a 2-h
recovery period, delayed memory tests were given. Results. Both doses
of midazolam decreased all scores in the memory and digit symbol subst
itution tests and mood ratings. Flumazenil reversed both the sedative
and the memory effects of the benzodiazepine. The reversal was as comp
lete with the 1-mg dose of flumazenil as with the 3-mg dose. Flumazeni
l by itself, and the acute reversal of midazolam effects, caused no si
gnificant behavioral reactions. Conclusions. Midazolam impairs explici
t and implicit memory. Flumazenil reverses both the sedative and memor
y effects of the drug. Flumazenil, in the doses used, has no intrinsic
actions.