Text-only Preview

2002, 77, 91–104
Pigeons were trained to discriminate 5 mg/kg pentobarbital from saline under concurrent vari-
able-ratio (VR) VR schedules, in which responses on the pentobarbital-biased lever were reinforced
under the VR schedule with the smaller response requirements when pentobarbital was given
before the session, and responses on the saline-biased key were reinforced under the VR schedule
with the larger response requirements. When saline was administered before the session, the re-
inforcement contingencies associated with the two response keys were reversed. When responding
stabilized under concurrent VR 20 VR 30, concurrent VR 10 VR 40, or concurrent VR 5 VR 50
schedules, pigeons responded almost exclusively on the key on which fewer responses were re-
quired to produce the reinforcer. When other doses of pentobarbital and other drugs were sub-
stituted for the training dose, low doses of all drugs produced responding on the saline-biased
key. Higher doses of pentobarbital and chlordiazepoxide produced responding only on the pen-
tobarbital-biased key, whereas higher doses of ethanol and phencyclidine produced responding
only on this key less often. d-Amphetamine produced responding primarily on the saline-biased
key. When drugs generalized to pentobarbital, the shape of the generalization curve under con-
current VR VR schedules was more often graded than quantal in shape. Thus, drug discrimination
can be established under concurrent VR VR schedules, but the shapes of drug-discrimination
dose–response curves under concurrent VR VR schedules more closely resemble those seen under
interval schedules than those seen under ?xed-ratio schedules. Graded dose–response curves un-
der concurrent VR VR schedules may relate to probability matching and dif?culty in discriminating
differences in reinforcement frequency.
Key words: drug discrimination, concurrent variable-ratio schedules, concurrent ?xed-ratio sched-
ules, pentobarbital, dose–response curves, key peck, pigeons
Most drug-discrimination research has con-
termine whether graded or quantal dose–re-
centrated on similarities and differences in
sponse curves are produced in stimulus gen-
the drugs that are used as discriminative stim-
eralization tests.
uli. The conditions under which drugs are es-
Although the in?uence of the schedule
tablished and maintained as discriminative
maintaining responding in drug discrimina-
stimuli have received much less attention.
tion may seem to be an esoteric question, it
One of the most important of these condi-
is fundamental to our understanding of stim-
tions is the schedule of reinforcement. Our
ulus generalization. As Bickel and Etzel
research has focused on the determination of
(1985) have pointed out, a quantal general-
how reinforcement contingencies in?uence
ization gradient implies that a stimulus func-
the shape of dose–response curves when oth-
tions as a unit that is either present or absent,
er drugs are substituted for the training drug
whereas a graded generalization gradient sug-
in drug-discrimination experiments. Speci?-
gests a proportional relation between stimu-
cally, we have addressed the question of
lus intensity and response strength. Quantal
whether drug-discrimination responses occur
generalization gradients, however, may some-
as quantal units or continuous variables un-
times appear to be continuous due to artifacts
der different reinforcement schedules. Our
of averaging all-or-nothing responding over
experimental approach has been to vary the
time or across subjects. For this reason, mean
reinforcement schedule systematically to de-
dose–response curves averaged across ani-
mals, or even across sessions, can be mislead-
This research was supported in part by Grant DA-02251
from the National Institute on Drug Abuse.
In drug-discrimination experiments, sub-
Reprints may be obtained from D. E. McMillan at the
jects usually are trained to make one re-
Department of Pharmacology and Toxicology, College of
sponse in the presence of a given dose of a
Medicine, University of Arkansas for Medical Sciences,
4301 West Markham Street, Little Rock, Arkansas 72205
training drug and a different response in its
(E-mail: [email protected]).
absence. Subsequently, generalization gradi-

D. E. MCMILLAN et al.
ents are determined by varying some dimen-
involve drug discrimination, Herrnstein and
sion of the discriminative stimulus, which in
Loveland (1975) and MacDonall (1988)
drug-discrimination experiments is the dose
have observed near-exclusive preference for
of the training drug. Under these conditions,
the smaller of a pair of VR schedules in a
most drug-discrimination generalization gra-
concurrent VR VR schedule when the VR
dients, at least in individual animals, have
components are independent of each other.
been quantal in form (Colpaert, 1991). In
However, there is reason to believe that
the great majority of experiments on drug
quantal responding does not always occur
discrimination, however, ?xed-ratio (FR)
under VR schedules in drug-discrimination
schedules of reinforcement have been used
experiments. Holloway and Gauvin (1989)
to maintain responding (Colpaert, 1987),
trained rats to discriminate 32 mg/kg caf-
and it has been suggested that the quantal
feine from saline under a VR schedule of re-
nature of drug-discrimination responding has
inforcement and found that graded re-
been imposed by the strict correlation be-
sponding occurred in individual subjects as
tween training conditions and the reinforce-
responding shifted from the saline key to the
ment of responding on only one of the levers
drug key as a function of increasing doses of
under an FR schedule (Colpaert, 1986,
caffeine. In the current experiments, we
studied pentobarbital discrimination in pi-
We have attempted to determine the role
geons maintained under several concurrent
of the reinforcement schedule in shaping
VR VR schedules to determine if the gener-
the drug-discrimination generalization gra-
alization gradient for pentobarbital and oth-
dient by studying drug discrimination under
er drugs was quantal (responding shifted
various schedules of reinforcement. The
from responding on the saline-biased key to
general ?nding has been that drug-discrim-
the pentobarbital-biased key without distrib-
ination generalization gradients are quantal
uting responses across both keys at inter-
when responding is maintained under FR
mediate doses), as would be predicted from
schedules and are graded when responding
the ?ndings of Herrnstein and Loveland
is maintained under ?xed-interval (FI) and
(1975), or if responding would be distribut-
variable-interval (VI) schedules of reinforce-
ed across both keys, as Holloway and Gauvin
ment. These effects occur under simple FI
(1989) observed with drug discriminations
schedules (Massey, McMillan, & Wessinger,
under simple VR schedules.
1992), simple FR schedules (Colpaert, 1987;
We selected several concurrent VR VR
Massey et al., 1992), VI schedules (Gouvier,
schedules for investigation, including concur-
Akins, & Trapold, 1984), second-order FR
rent VR 5 VR 50, concurrent VR 10 VR 40,
schedules (McMillan, Cole-Fullenwider,
and concurrent VR 20 VR 30. The concur-
Hardwick, & Wenger, 1982), multiple FR FI
rent VR 10 VR 40 schedule was chosen for
schedules (McMillan & Hardwick, 1996;
the initial schedule because we have used sim-
Snodgrass & McMillan, 1991), concurrent
ilar schedule values to maintain drug discrim-
FR FR schedules (McMillan & Li, 1999a),
ination under concurrent FR FR schedules
concurrent FI FI schedules (McMillan & Li,
(McMillan & Li, 1999a). When we did not
2000; McMillan, Li, & Hardwick, 1997), and
?nd the results that we expected under this
concurrent VI VI schedules (Snodgrass &
schedule, the other two concurrent schedules
McMillan, 1996).
were studied. These other schedule values
Conspicuous for their absence in this list
were chosen because under the concurrent
of schedules are concurrent variable-ratio
VR 5 VR 50 schedule, the ratio of responses
(VR) VR schedules. If responding under VR
required to produce the reinforcer under the
schedules is similar to responding under FR
two components of the schedule is 10:1, a val-
schedules, it would be anticipated that drug-
ue that should favor a preponderance of re-
discrimination generalization gradients de-
sponding on the response key associated with
termined under concurrent VR VR sched-
the VR 5 schedule component. In contrast,
ules would generate quantal dose–response
under the concurrent VR 20 VR 30 this ratio
curves, as has been shown to occur with con-
is only 1.5:1, which might result in a more
current FR FR schedules (McMillan & Li,
even distribution of responses across the two
1999a). Indeed, in experiments that did not
keys. The drugs chosen for the stimulus gen-

eralization determinations were the same as
those that we have used in most of our pre-
The training of pigeons to be used in ex-
vious experiments.
periments on drug discrimination under con-
current schedules has been described in de-
tail previously (Snodgrass & McMillan, 1996).
Brie?y, they were trained to peck the two
lighted side keys on the response panel by
autoshaping. After the pigeons had earned
Five male White Carneau pigeons, weigh-
50 reinforcers for responding on each key
ing between 414 and 522 g at 80% of their
with only one key lighted, both keys were
free-feeding weights, were used as subjects for
lighted and the schedule was changed to con-
the experiment. They were maintained at
current VR 2 VR 8. During alternating ses-
these weights by food earned during experi-
sions, 5 mg/kg pentobarbital or saline was ad-
mental sessions and by supplemental feeding
ministered 10 min before the session. If
immediately after the sessions. The birds
pentobarbital was administered before the
were experimentally naive at the beginning of
session, responding on the left key was rein-
these experiments. They were maintained in
forced under the VR 2 component and re-
a colony room in which temperature and hu-
sponding on the right key was reinforced un-
midity were controlled. Lights were on in the
der the VR 8 component. If saline was
colony room from 7:00 a.m. to 7:00 p.m. Wa-
administered before the session, the rein-
ter and grit were freely available in the home
forcement schedules were reversed on the
cages, but not in the test cages.
two keys. Over several sessions, the sizes of
the VR components were increased to their
?nal values. Training sessions continued until
Test sessions were conducted in a Ger-
44 reinforcers had been delivered or until
brands pigeon chamber (Model G-5610) en-
2,400 s had elapsed, whichever occurred ?rst.
closed in a Gerbrands sound- and light-atten-
Pigeons were maintained under each concur-
uating cubicle (Model G-7211). On the front
rent schedule until there were no increasing
panel of the cage, three Gerbrands response
or decreasing trends in the ratio of responses
keys (Model 7410) were mounted 7 cm apart,
on the two keys over six consecutive training
20 cm above the grid ?oor. A force of 15 g
sessions. Hereafter, the key on which respons-
was required to operate each key. Keys could
es were reinforced under the VR schedule
be transilluminated with various colored
component with the lower VR requirement
lights, but the center key was not used in
after pentobarbital administration will be re-
these experiments and it remained dark at all
ferred to as the pentobabital-biased key, and
times. During these experiments, the left key
the key on which responses were reinforced
was transilluminated with a green light and
under the schedule with the lowest VR re-
the right key with a red light for Pigeons 386,
quirement after saline administration will be
388, and 390. The key colors were reversed
referred to as the saline-biased key.
for Pigeons 387 and 389. The reinforcer was
After responding stablized, dose–response
4-s access to mixed grain, presented by a food
curves for pentobarbital and other drugs
hopper through an opening (6 cm square)
were determined during the next several
located 2 cm above the grid ?oor centered
months, after which the schedule was
below the keys. Two 28-VDC lights illuminat-
changed to a new concurrent VR VR sched-
ed the food hopper when it was operated. A
ule and training continued under the new
28-VDC houselight mounted on the front left
schedule until performance again stabilized
corner of the top panel illuminated the
(about 25 training sessions); then the effects
chamber when schedule contingencies were
of the drugs were redetermined. This process
in effect. Experiments were controlled and
continued until the effects of drugs on per-
data recorded by a Gateway 2000 microcom-
formance under all three concurrent VR VR
puter through a Microcomputer Interface II
schedules had been determined. All pigeons
(Med Associates, Inc.) using software devel-
were trained under the concurrent VR 10 VR
oped in our laboratory.
40 ?rst. After completion of the dose–re-

D. E. MCMILLAN et al.
sponse curves under concurrent VR 10 VR
Data Analysis
40, Pigeons 386 and 387 were switched to a
The number of responses made on each
concurrent VR 20 VR 30 schedule and Pi-
key and the number of reinforcers produced
geons 388, 389, and 390 were switched to a
by responses on each key were recorded and
concurrent VR 5 VR 50 schedule. After com-
reported as percentages of total responses or
pletion of the dose–response curves with re-
total reinforcers. Data from single observa-
sponding maintained under these schedules,
tions in individual animals that were obtained
the schedules were reversed for the two
by varying the dose of each drug were plotted
groups of pigeons, so that all pigeons were
as a percentage of responses on the pento-
exposed to all three concurrent VR VR sched-
barbital-biased key for graphic analysis. Over-
all rates of responding were calculated, and
The effects of pentobarbital, d-amphet-
the number of changes from responding on
amine, chlordiazepoxide, ethanol, and phen-
one key to responding on the other (change-
cyclidine on responding under the concurrent
overs; COs) were recorded for each pigeon
VR 10 VR 40 schedule were studied in that
and were compared with means during train-
order, and the doses of each drug were given
ing sessions with saline and 5 mg/kg pento-
in a mixed order. After responding stabilized
under the second concurrent schedule to
De?ning what is a quantal dose–response
which the birds were exposed, the order of
curve and what is a graded dose–response
drug exposure was changed to pentobarbital,
curve is somewhat arbitrary. As is commonly
phencyclidine, chlordiazepoxide, d-amphet-
done in drug-discrimination research, sub-
amine, and ethanol. After responding stabi-
jects were considered to have made the sa-
lized under the ?nal concurrent schedule to
line-biased response if less than 20% of re-
which the pigeons were exposed, the order of
sponses occurred on the pentobarbital-biased
drug exposure was changed to pentobarbital,
key after a drug administration. They were
phencyclidine, d-amphetamine, chlordiaz-
considered to have made the pentobarbital-
epoxide, and ethanol. After responding stabi-
biased response if more than 80% of re-
lized under a given schedule, test drugs were
sponses occurred on the pentobarbital-bi-
given on Tuesdays and Fridays. All dose–re-
ased key after a drug administration.
sponse curves were based on single observa-
Individual dose–response curves were con-
tions in each pigeon. Additional training ses-
sidered to be quantal if no points on the
dose–response curve were between 20% and
Wednesdays, and Thursdays.
80% on the pentobarbital-biased key. Dose–
response curves were considered to be grad-
ed if at least one point on the dose–response
Injections were administered into the
curve was between 20% and 80% of respons-
breast muscle 10 min before the 40-min ses-
es on the pentobarbital-biased key. The fre-
sion in a volume of 0.1 ml/100 g of body
quency with which quantal and graded dose–
weight, with the exception of ethanol, which
response cur ves were produced under
was administered into the proventriculus by a
different concurrent VR VR schedules and
gavage needle 15 min before the session. The
under different orders of exposure to these
drugs studied were sodium pentobarbital
schedules was tested for statistical signi?-
(Sigma Chemical Co., St. Louis, MO), d-am-
cance by chi-square.
phetamine hydrochloride (Sigma Chemical
Co.), chlordiazepoxide hydrochloride (kindly
supplied by Hoffmann La Roche, Nutley, NJ),
and phencyclidine hydrochloride (kindly sup-
Table 1 shows performance of individual
plied by NIDA, Rockville, MD). Drugs were
pigeons for the last six training sessions after
dissolved in physiologic saline and doses were
saline administration and the last six sessions
calculated as the salts, except for ethanol,
after pentobarbital administration during ses-
which was obtained from the University Hos-
sions conducted before drug-substitution
pital as a 100% solution and diluted to a 10%
tests were initiated in pigeons trained under
(wt/vol) solution with tap water.
the concurrent VR VR schedules. Stimulus

Table 1
Mean and standard deviation (in parentheses) of the percentage of total responses on the
saline-biased key and the pentobarbital-biased key, percentage of total reinforcers delivered
following responses on these keys, overall response rate (responses per second) and change-
overs (CO) during six saline and six pentobarbital training sessions for individual animals
maintained under each concurrent schedule.
Saline-training sessions
Pentobarbital-training sessions
% saline key
% pentobarbital key
Bird Responses Reinforcers
Responses Reinforcers
98 (4)
99 (2)
1.54 (0.23)
2.6 (2.5)
97 (6)
98 (5)
1.79 (0.07)
0.3 (0.8)
VR 20 VR 30
95 (13)
96 (10)
1.91 (0.04)
0.1 (0.4)
97 (3)
97 (3)
2.13 (0.03)
0.9 (0.4)
100 (0)
100 (0)
2.02 (0.15)
0.0 (0)
94 (4)
95 (3)
1.82 (0.04)
2.6 (2.5)
96 (4)
97 (6)
2.05 (0.10)
0.4 (0.8)
100 (0)
99 (1)
2.16 (0.08)
0.3 (0.5)
100 (0)
100 (0)
1.86 (0.11)
0.1 (0.4)
95 (10)
100 (0)
1.79 (0.28)
0.1 (0.4)
96 (4)
98 (2)
1.42 (0.07)
1.0 (1)
100 (0)
100 (0)
1.40 (0.04)
0.0 (0)
VR 10 VR 40
99 (1)
99 (1)
1.40 (0.03)
0.3 (0.7)
96 (4)
98 (1)
1.60 (0.04)
1.0 (0.5)
100 (0)
100 (0)
1.52 (0.17)
0.1 (0.3)
90 (17)
95 (7)
1.42 (0.17)
2.1 (2.7)
97 (6)
99 (2)
1.59 (0.08)
0.4 (1)
98 (6)
97 (2)
1.64 (0.06)
1.0 (0)
100 (0)
100 (0)
1.42 (0.13)
0.7 (2)
98 (4)
99 (2)
1.42 (0.04)
0.4 (0.7)
99 (2)
100 (0)
0.92 (0.03)
0.1 (0.4)
99 (2)
99 (1)
0.83 (0.11)
0.1 (0.4)
VR 5 VR 50
92 (7)
97 (2)
1.10 (0.04)
0.7 (0.5)
100 (0)
100 (0)
0.92 (0.02)
0.7 (0.5)
95 (2)
97 (1)
0.87 (0.04)
0.9 (0.4)
100 (0)
100 (0)
0.84 (0.05)
0.9 (0.4)
97 (3)
98 (1)
1.04 (0.08)
1.4 (1.9)
98 (2)
99 (1)
0.96 (0.05)
1.4 (1.9)
100 (0)
100 (0)
0.91 (0.03)
0.0 (0)
100 (0)
100 (0)
0.88 (0.05)
0.0 (0)
control by the presence or absence of pen-
responses per second, 0.07 SEM, after saline
tobarbital was strong under all three concur-
and 1.94 responses per second, 0.17 SEM, af-
rent schedules. After administration of saline,
ter pentobarbital). Overall rates of respond-
the pigeons averaged 98%, 98%, and 97% of
ing were slightly lower under the concurrent
their responses on the saline-biased key un-
VR 10 VR 40 schedule (1.47 responses per
der the concurrent VR 20 VR 30, concurrent
second, 0.07 SEM, after saline and 1.50 re-
VR 10 VR 40, and concurrent VR 5 VR 50
sponses per second 0.10 SEM, after pentobar-
schedules, respectively. After administration
bital) and were much lower under the con-
of pentobarbital, the percentages of respons-
current VR 5 VR 50 schedule (0.97 responses
es on the pentobarbital-biased key were 97%,
per second 0.09 SEM, after saline and 0.89
96%, and 99% for these same three concur-
responses per second, 0.05 SEM, after pen-
rent schedules. The percentage of responses
on each key was closely approximated by the
Figure 1 shows the dose–response curves
percentage of reinforcers delivered for re-
for the effects of increasing doses of pento-
sponding on that key, as would be expected
barbital on responding under each of the
under a ratio schedule. Thus, stimulus con-
three concurrent schedules for individual pi-
trol did not differ depending on whether sa-
geons. In general, as the dose of pentobar-
line or pentobarbital was administered, or on
bital increased, responding moved from the
the values of the concurrent schedule. The
saline-biased key to the pentobarbital-biased
number of COs was low for all pigeons, av-
key. After the 10 mg/kg dose, all pigeons re-
eraging less than one per training session af-
sponded almost exclusively on the pentobar-
ter both saline and pentobarbital under all
bital key under all three concurrent sched-
three concurrent schedules. Rates of re-
ules. Both quantal and graded dose–
sponding were similar after saline and pen-
response curves were observed in individual
tobarbital administration; however, there
pigeons. Graded dose–response curves oc-
were differences in rates of responding that
curred for Pigeons 386, 388, and 389 under
depended on the schedule. The highest base-
the concurrent VR 20 VR 30 schedule; Pi-
line rates of responding were observed under
geons 386, 387, and 390 under the concur-
the concurrent VR 20 VR 30 schedule (1.88
rent VR 10 VR 40 schedule; and Pigeons 386,

D. E. MCMILLAN et al.
Fig. 2.
Drug-discrimination generalization gradients
Fig. 1.
Drug-discrimination generalization gradients
for chlordiazepoxide under concurrent VR 20 VR 30
for pentobarbital under concurrent VR 20 VR 30 (top),
(top), concurrent VR 10 VR 40 (middle), and concurrent
concurrent VR 10 VR 40 (middle), and concurrent VR 5
VR 5 VR 50 (bottom). Details as in Figure 1.
VR 50 (bottom). Abscissa: milligrams per kilogram dose
on a log scale. Ordinate: percentage of responses on pen-
tobarbital-biased key. Symbols and brackets at CONTROL
DATA show means and standard deviations for six training
sessions after responding stabilized. The top set of brackets
and symbols is for training sessions after pentobarbital was
387, 388, and 390 under the concurrent VR
administered, and the lower set of brackets and symbols is
5 VR 50 schedule.
for training sessions after saline was administered. If no
Figure 2 shows dose–response curves for
brackets are shown, standard deviations are smaller than
the effects of chlordiazepoxide as a discrimi-
the symbol. Symbols for dose–response curves represent
native stimulus under each of the concurrent
single observations in individual pigeons. Filled circles, Pi-
geon 386; open circles, Pigeon 387; ?lled triangles, Pigeon
schedules. Under the concurrent VR 20 VR
388; open triangles, Pigeon 389; ?lled squares, Pigeon 390.
30 schedule, only 3 pigeons (386, 387, and

389) responded predominantly on the pen-
tobarbital-biased key after high doses of
chlordiazepoxide. The other 2 responded
largely on the saline-biased key. Only the data
from Pigeon 386 met the criterion for a grad-
ed dose–response curve. Under the concur-
rent VR 10 VR 40 schedule, all pigeons ex-
cept 389 responded predominantly on the
pentobarbital-biased key after high doses of
chlordiazepoxide, and even this pigeon made
about half of its responses on the pentobar-
bital-biased key after 10 mg/kg chlordiaz-
epoxide. The dose–response curves for every
pigeon met the criterion for being graded,
although some of the curves showed irregu-
lar reversals in the effects of some doses. Un-
der the concurrent VR 5 VR 50 schedule, all
pigeons except 388 responded predominant-
ly on the pentobarbital-biased key after high
doses of chlordiazepoxide, and all pigeons
except 389 exhibited graded dose–response
curves. Thus, the pentobarbital stimulus gen-
eralized to higher doses of chlordiazepoxide
in most pigeons under most schedules, and
the shape of the dose–response curves was
more often graded than quantal.
Figure 3 shows the dose–response curves
for ethanol as a discriminative stimulus under
the three concurrent schedules. Under con-
current VR 20 VR 30, only Pigeons 386 and
389 responded on the pentobarbital-biased
key after ethanol. The dose–response curve
was graded only for Pigeon 386. Under con-
current VR 10 VR 40, only Pigeon 388 failed
to respond predominantly on the pentobar-
bital-biased key after some dose of ethanol,
although for Pigeon 390 there was a dosage
reversal of effect between 1,000 and 1,800
mg/kg. Only the dose–response curves for Pi-
geons 386 and 387 were graded. Under con-
current VR 5 VR 50, only with Pigeon 390 was
Fig. 3.
Drug-discrimination generalization gradients
responding con?ned to the pentobarbital key
for ethanol under concurrent VR 20 VR 30 (top), con-
after high doses of ethanol. To the extent that
current VR 10 VR 40 (middle), and concurrent VR 5 VR
differential responding occurred with Pi-
50 (bottom). Details as in Figure 1.
geons 386, 389, and 390, dose–response
curves were graded. With Pigeons 387 and
388, responding was con?ned largely to the
saline-biased key.
doses of phencyclidine. The dose–response
Figure 4 shows the effects of phencycli-
curves under this schedule met the criteri-
dine as a discriminative stimulus under the
on for quantal dose–response curves for 2
concurrent schedules. Under concurrent
of these pigeons (389 and 390). Pigeon 388
VR 20 VR 30, 4 of the pigeons (386, 387,
responded only on the saline key until a
389, and 390) responded predominantly on
dose was reached that eliminated respond-
the pentobarbital-biased key after higher
ing. Under concurrent VR 10 VR 40, the

D. E. MCMILLAN et al.
rent VR 5 VR 50, more than 80% of the
responses occurred on the pentobarbital-bi-
ased key after phencyclidine in 2 pigeons
(387 and 389), and Pigeon 386 approached
this level. The other 2 pigeons responded
only on the saline key at doses that did not
eliminate responding. The dose–response
curve for phencyclidine was graded in 2 of
the 3 pigeons (386 and 389) that emitted a
substantial amount of responding on the
pentobarbital-biased key, and was quantal in
the 3rd (387).
Figure 5 shows the effects of d-amphet-
amine as a discriminative stimulus under the
three concurrent schedules. Under concur-
rent VR 20 VR 30, none of the pigeons re-
sponded predominantly on the pentobarbi-
tal-biased key, although Pigeons 386, 389, and
perhaps 387 met the criterion for graded re-
sponding across the two keys after some dos-
es. Under concurrent VR 10 VR 40, similar
effects were observed, with considerable re-
sponding occurring on both keys for Pigeons
386, 387, and to a lesser extent, 390. Re-
sponding by the other pigeons was con?ned
to the saline-biased key. Under concurrent
VR 5 VR 50, responding was generally con-
?ned to the saline key, except that for Pigeon
386 responding occurred on both keys, es-
pecially after the highest dose. Under all
three schedules, these curves with interme-
diate effects were sometimes characterized by
irregular dose–response curves.
To determine if there were differences
among schedules in the shapes of the dose–
response curves that were generated under
the different concurrent VR VR schedules, a
chi-square analysis was conducted across all
drugs using only the dose–response curves
that met criteria for full generalization to the
training drug. The frequency of graded and
Fig. 4.
Drug-discrimination generalization gradients
quantal dose–response curves did not differ
for phencyclidine under concurrent VR 20 VR 30 (top),
concurrent VR 10 VR 40 (middle), and concurrent VR 5
signi?cantly across the three concurrent VR
VR 50 (bottom). Details as in Figure 1.
VR schedules. Similarly, chi-square tests were
conducted to determine if the order of ex-
posure to the concurrent VR 20 VR 30 and
the concurrent VR 5 VR 50 schedules in?u-
results were similar, with the same 4 pigeons
enced the frequency of graded and quantal
responding predominantly on the pento-
dose–response curves under these schedules.
barbital-biased key after phencyclidine and
Again, the results of these tests were not sta-
Pigeon 388 responding only on the saline
tistically signi?cant.
key. The dose–response curves met the cri-
Table 2 shows the effects of each of the
terion for being graded for all of these pi-
test drugs on overall rates of responding.
geons except Pigeon 390. Under concur-
Rates of responding increased after each test

Table 2
Effects of drugs on rates of responding (responses per
second) under the three concurrent VR VR schedules of
reinforcement. Each value represents a mean from single
observations in 5 pigeons. ? indicates that rates were
more than two standard deviations below the control
mean on saline-training days. ? indicates that the rates
were more than two standard deviations above the con-
trol mean on saline-training days.
reinforcement schedule
VR 20
VR 10
VR 5
VR 30
VR 40
VR 50
Note. NT
dose not tested.
creased overall rates of responding under
concurrent VR 10 VR 40. Under concurrent
VR 20 VR 30, high doses of all drugs except
pentobarbital decreased rates of responding.
The rate decreases after phencyclidine were
marked compared to those of the other
Fig. 5.
Drug-discrimination generalization gradients
for d-amphetamine under concurrent VR 20 VR 30 (top),
concurrent VR 10 VR 40 (middle), and concurrent VR 5
VR 50 (bottom). Details as in Figure 1.
The presence or absence of the training
drug (relative to saline control sessions) for
dose of pentobarbital precisely controlled the
sessions under concurrent VR 5 VR 50. Un-
location of responding under concurrent VR
der concurrent VR 10 VR 40, pentobarbital
VR schedules of reinforcement after respond-
had little effect on rates of responding at the
ing had stabilized. Averaged across pentobar-
doses studied. Low doses of ethanol in-
bital and saline training conditions, across
creased rates of responding under this
the three concurrent VR VR schedules, and
schedule, but the highest dose decreased re-
across individual pigeons, 98.5% (1.5% SEM)
sponse rates. High doses of other drugs de-
of responses occurred on the key with the

D. E. MCMILLAN et al.
lower ratio requirement (Table 1). Thus, the
sometimes been reported for other rein-
location of responding came under the con-
forcement schedules (McMillan & Li,
trol of both the drug condition and the VR
response requirement. We had anticipated
Because previous studies were consistent in
that as the difference between the size of the
showing graded dose–response curves under
VR components decreased, more responding
interval schedules and quantal dose–response
would occur on the key with the higher ratio
curves under ratio schedules, it seemed pos-
value than would occur with larger differenc-
sible that the shape of the curves depended
es in the size of the VR components, but this
on whether the schedule was a ratio or an
was not the case. The distribution of respons-
interval schedule. In a previous experiment
es across the two keys appeared to be inde-
with concurrent FR 10 FR 40 schedules, pi-
pendent of the differences in the VR values
geons trained to discriminate 5 mg/kg pen-
across the range of values studied in this ex-
tobarbital from saline exhibited quantal
periment. Other investigators also have ob-
dose–response curves for pentobarbital and
served exclusive responding on the manipu-
other drugs that generalize to pentobarbital
landum associated with the lower ratio value
(McMillan & Li, 1999a). The ratio schedule
under concurrent VR VR schedules in exper-
values in that study were identical to the con-
iments that did not involve drug discrimina-
current VR 10 VR 40 schedules in the present
tions (Herrnstein & Loveland, 1975; Mac-
study, yet under the concurrent FR FR sched-
Donall, 1988).
ule almost all of the dose–response curves
The major reason for conducting the pres-
were quantal, whereas under the concurrent
ent experiment was to determine the shape
VR VR schedule in the present study most
of the drug-discrimination generalization
were graded. The much more frequent oc-
gradients under concurrent VR VR sched-
currence of graded dose–response curves un-
ules. In a long series of experiments, we have
der the concurrent VR VR schedule than un-
been building a case that interval schedules
der the concurrent FR FR schedule suggests
generate graded dose–response cur ves,
that factors other than ratio or interval sched-
whereas ratio schedules generate quantal
uling of the delivery of the reinforcer play a
dose–response cur ves (see the introduc-
role in the determination of the shape of the
tion); however, all of our previous drug-dis-
generalization dose–response curves.
crimination experiments with ratio sched-
A possible explanation for the differences
ules had been conducted with some type of
in the shapes of the dose–response curves
FR schedule (Massey et al., 1992; McMillan
under concurrent VR VR schedules and con-
& Hardwick, 1996; McMillan & Li, 1999a).
current FR FR schedules may be differences
Obser vations by Holloway and Gauvin
in the discriminability of reinforcer rates un-
(1989) had suggested that drug-discrimina-
der VR and FR schedules. Under some con-
tion generalization gradients under VR
ditions, probability matching can occur un-
schedules might not have the same shape as
der concurrent ratio schedules, whereby the
those under FR schedules. This observation
animal matches the ratio of alternative re-
was con?rmed in the present experiments,
sponses to the probability of reinforcement
in that the majority of generalization gradi-
for each alternative, even though a higher
ents for those drugs that generalized com-
reinforcement rate would occur if the ani-
pletely to the training dose of pentobarbital
mal responded exclusively on the manipu-
generated graded rather than quantal dose–
landum with the higher reinforcement rate
response curves under concurrent VR VR
(Bitterman, 1965). Clearly, probability
schedules. Chi-square tests suggested that
matching did not occur under baseline con-
there was no difference among the three
ditions in the present study, because re-
schedules of reinforcement in the frequency
sponding was con?ned almost exclusively to
with which graded and quantal dose–re-
the alternative with the highest reinforce-
sponse curves were generated across drugs,
ment rate under the concurrent VR VR
nor were there statistically signi?cant differ-
schedule. Nevertheless, it is possible, under
ences in the frequency of graded and quan-
the in?uence of a drug dose that is not the
tal dose–response curves relating to the or-
training dose or training drug, that proba-
der of exposure to the schedules, as has
bility matching may occur under concurrent