JOURNAL
OF
THE
EXPERIMENTAL
ANALYSIS
OF
BEHAVIOR
EFFECTS
OF
INCOME
ON
DRUG
CHOICE
IN
HUMANS
R.
J.
DEGRANDPRE,
WARREN
K.
BICKEL,
S.
ABU
TURAB
Rizvi,
AND
JOHN
R.
HUGHES
UNIVERSITY
OF
VERMONT
The
effects
of
income
(money
available
to
spend
during
the
experimental
session)
on
human
choice
were
examined
in
a
concurrent-schedule
arrangement.
Subjects
were
7
nicotine-dependent
smokers,
and
reinforcers
were
puffs
on
the
subject's
usual
brand
of
cigarette
("own")
and
puffs
on
a
less
preferred
brand
of
cigarette
with
equal
nicotine
content
("other").
Across
sessions,
income
varied
and
the
price
of
the
two
reinforcers
was
held
constant,
with
the
other
puffs
one
fifth
the
price
of
the
own
puffs.
As
income
increased,
consumption
of
own
puffs
increased
while
consumption
of
the
less
expensive
other
puffs
decreased.
These
effects
of
income
on
choice
were
highly
consistent
across
subjects.
For
some
subjects,
however,
income
had
little
effect
on
total
puff
consumption.
Finally,
an
additional
condition
examined
whether
price
and
income
manipulations
would
have
functionally
equivalent
effects
on
choice
by
repeating
an
income
condition
in
which
the
price
of
the
other
brand
was
increased.
Although
the
increased
price
of
the
other
puffs
decreased
their
consumption
in
4
subjects,
2
subjects
showed
increased
consumption
of
the
other
puffs
at
the
higher
price.
The
results,
when
defined
in
economic
terms,
indicate
that
the
own
puffs
were
a
normal
good
(consumption
and
income
are
directly
related),
the
other
puffs
were
an
inferior
good
(consumption
and
income
are
inversely
related),
and
the
direct
relationship
between
consumption
of
the
other
puffs
and
their
price
is
defined
as
a
Giffen-
good
effect.
The
latter
result
also
suggests
that
for
these
2
subjects,
price
and
income
manipulations
had
equivalent
effects
on
choice.
These
results
extend
findings
from
previous
studies
that
have
examined
the
effects
of
income
on
choice
responding
to
human
subjects
and
drug
reinforcers,
and
provide
a
framework
for
further
experimental
tests
of
the
effects
of
income
on
human
choice
behavior.
Meth-
odological
and
theoretical
implications
for
the
study
of
choice
and
for
behavioral
pharmacology
are
discussed.
Key
words:
behavioral
economics,
behavioral
pharmacology,
choice,
demand,
drug
reinforcers,
nic-
otine,
drug
policy,
income,
lever
press,
humans
Numerous
studies
have
investigated
choice
between
drug
reinforcers
and
between
drug
and
nondrug
reinforcers.
Under
these
proce-
dures,
two
or
more
reinforcers
are
available
in
a
concurrent-schedule
arrangement.
Respond-
ing
on
one
manipulandum
is
maintained
by
drug
reinforcement,
while
responding
on
an-
other
manipulandum
is
maintained
by
either
a
different
dose
of
the
same
drug,
a
different
drug,
or
a
nondrug
reinforcer.
These
studies
are
typically
used
to
assess
the
reinforcing
ef-
fectiveness
of
one
stimulus
relative
to
some
This
research
was
conducted
when
the
first
author
was
a
predoctoral
student
in
the
Department
of
Psychology,
University
of
Vermont.
This
work
is
funded
by
a
Sigma
Xi
Grant-in-aid
(R.J.D.),
National
Institute
on
Drug
Abuse
Training
Grant
T32
DA-07242
(R.J.D.),
and
Grants
RO1
DA-06626
(W.K.B.),
and
PHS
K02-00109
(J.R.H.).
We
thank
Melissa
Jurgens
and
Michael
Layng
for
their
assistance
in
preparing
the
manuscript.
Send
requests
for
reprints
to
the
first
or
second
author
at
the
Department
of
Psychiatry,
Human
Behavioral
Pharma-
cology
Laboratory,
38
Fletcher
Place,
University
of
Ver-
mont,
Burlington,
Vermont
05401-1419.
other
stimulus.
As
stated
by
Katz
(1990),
the
relative
frequency
of
responding
for
one
re-
inforcer
"is
used
as
an
indication
of
the
dif-
ferences
in
reinforcing
effectiveness
of
the
two
consequences"
(p.
285).
For
example,
one
study
examined
choice
responding
for
different
doses
of
intravenous
cocaine
in
rhesus
monkeys
using
concurrent
variable-interval
1-min
schedules
(Iglauer
&
Woods,
1974).
Greater
responding
(and
some-
times
exclusive
responding)
occurred
for
the
higher
dose,
suggesting
greater
effectiveness
of
the
higher
doses.
Other
studies
have
examined
choice
between
different
drug
reinforcers.
For
example,
one
study
examined
the
effects
of
methylphenidate
dose
on
responding
main-
tained
by
it
and
one
of
two
doses
of
cocaine
(Johanson
&
Schuster,
1975).
Preference
for
cocaine
decreased
as
the
comparison
dose
of
methylphenidate
increased,
suggesting,
again,
that
the
effectiveness
of
a
reinforcer
may
vary
as
a
function
of
dose.
These
and
numerous
other
studies
have
shown
the
utility
of
choice
procedures
for
examining
relative
reinforcing
483
NUMBER
3
(MAY)
1993,
59,
483-500
R.
J.
DeGRANDPRE
et
al.
effectiveness
of
pharmacological
and
non-
pharmacological
substances
(for
a
review
of
studies
on
choice
and
reinforcer
effectiveness,
see
Katz,
1990).
Other
studies
suggest
that
choice
may
also
be
a
function
of
variables
that
are
independent
of
the
reinforcers
(e.g.,
intertrial
interval;
ITI).
One
such
study
examined
whether
choice
be-
tween
heroin
and
food,
at
constant
magnitudes,
would
vary
as
a
function
of
ITI
(Elsmore,
Fletcher,
Conrad,
&
Sodetz,
1980).
Even
though
an
equal
number
of
choices
for
heroin
and
food
occurred
at
low
ITIs,
heroin
choices
decreased
considerably
more
than
food
choices
as
ITI
increased.
This
finding
is
important
in
showing
that
choice
between
two
reinforcers
can
be
altered
by
variables
that
are
indepen-
dent
of
the
schedule
and
magnitude
of
rein-
forcement
for
the
concurrently
available
re-
inforcers.
More
recently,
the
manipulation
made
in
the
Elsmore
et
al.
study
(1980)
has
been
conceptualized
in
terms
of
the
economic
notion
of
income
and
has
been
further
shown
to
affect
choice
between
concurrently
available
different
reinforcers
(Silberberg,
Warren-
Boulton,
&
Asano,
1987).
Income
can
be
defined
as
the
amount
of
funds,
goods,
or
services
available
to
any
one
indi-
vidual
at
any
given
time
(Pearce,
1986).
In
behavioral
terms,
income
manipulations
can
be
conceptualized
as
constraints
on
total
re-
inforcement
within
a
session.
Microeconomic
theory
considers
the
relation
between
income
and
consumption
and
recognizes
that
increases
in
income
can
either
increase
or
decrease
choice
of
any
particular
good,
depending
on
the
type
of
good
and
the
availability
of
other
goods
(Deaton
&
Muellbauer,
1980;
Lea,
Tarpy,
&
Webley,
1987).
For
example,
an
increase
in
income
might
increase
seafood
consumption
while
decreasing
consumption
of
hamburger.
The
former
relationship
is
defined
in
economic
terms
as
a
normal
good
(i.e.,
normal-good
ef-
fect),
and
the
latter
is
defined
as
an
inferior
good
(i.e.,
inferior-good
effect).
Normal
goods
often
correspond
to
the
everyday
notion
of
lux-
uries,
whereas
inferior
goods
are
often
neces-
sities.
In
recent
experimental
analyses,
the
eco-
nomic
notion
of
income
has
been
extended
to
nonhuman
choice
between
nonidentical
rein-
forcers
(Hastjarjo
&
Silberberg,
1992;
Hast-
jarjo,
Silberberg,
&
Hursh,
1990;
Silberberg
et
al.,
1987).
Consistent
with
the
definitions
above,
studies
investigating
income
have
de-
fined
income
as
the resources
available
to
the
organism
to
obtain
food
during
the
experi-
mental
session
(i.e.,
number
of
trials
or
rein-
forcers
per
session;
Battalio,
Kagel,
&
Kogut,
1991;
Hastjarjo
&
Silberberg,
1992;
Hastjarjo
et
al.,
1990).
For
example,
in
one
study
of
macaque
monkeys,
ITI
was
varied
as
the
in-
come
manipulation
(i.e.,
number
of
reinforced
trials
per
session)
while
session
length
was
held
constant
(Silberberg
et
al.,
1987,
Exper-
iment
1).
Choice
of
a
large
bitter-tasting
food
pellet
increased
relative
to
a
small
normal
pel-
let
when
income
was
decreased.
In
economic
terms,
the
bitter
pellet
was
an
inferior
good
and
the
small
normal
pellet
was
a
normal
good
in
this
context
(Silberberg
et
al.,
1987).
A
sec-
ond
study
manipulating
income
(Hastjarjo
et
al.,
1990)
replicated
the
effects
of
income
re-
ported
in
the
Silberberg
et
al.
study
using
rats
as
subjects,
and
a
third
study
(Hastjarjo
&
Silberberg,
1992)
extended
these
findings
to
show
that
income
could
significantly
alter
choice
between
an
immediate
and
delayed
re-
inforcer
in
rats.
The
present
study
assessed
the
effect
of
income
on
drug
choice
in
human
cig-
arette
smoking.
This
study
also
examined,
in
a
more
limited
scope,
whether
these
income
manipulations
might
have
effects
on
drug
choice
that
are
functionally
equivalent
to
those
of
price
manipulations.
Although
considerable
research
has
been
conducted
on
choice
using
drug
reinforcers,
to
our
knowledge
there
are
no
laboratory
studies
that
explicitly
manipulated
income
using
drugs
as
reinforcers.
Nor,
to
our
knowledge,
have
the
effects
of
income
been
investigated
in
hu-
mans
in
a
laboratory
setting.
METHOD
Subjects,
Apparatus,
and
Reinforcers
One
female
and
6
male
smokers
partici-
pated.
Subjects
ranged
from
20
to
38
years
old
and
smoked
one
or
more
packs
of
0.6-
to
1.2-
mg
nicotine
cigarettes
per
day.
Nicotine
self-
administration
via
cigarette
smoking
was
used
in
this
study
because
it
is
well
established
as
a
potent
reinforcer
in
humans
(Henningfield
&
Goldberg,
1983)
and
because
it
has
many
features
of
other
drugs
of
dependence
(West
&
Grunberg,
1991)
that
are
responsive
to
en-
484
EFFECTS
OF
INCOME
ON
CHOICE
Table
1
Subject
characteristics.
Cigarettes
Fagerstrom
per
day
Nicotine/tar
Subject
Age
Gender
scores
(avg.)
Preferred
brand
(mg/cigarette)
BT
35
M
8
25
American
Lightss
1.2/16
JH
38
M
9
30
Winstons
1.1/17
JR
22
M
8
30
Marlboro,
1.2/17
KC
20
F
7
20
Camel
LightsD
0.7/09
KS
36
M
9
30
Merits
0.6/08
PZ
25
M
7
25
Camel1
filter
1.0/15
WR
20
M
8
20
Marlboros
1.2/17
vironmental
variations
(Hughes,
1989).
Sub-
jects
were
recruited
from
newspaper
adver-
tisements,
were
in
good
health,
and
reported
no
medication
usage
or
drug
or
alcohol
abuse
other
than
nicotine.
Subjects
had
to
score
a
minimum
of
7
on
the
Fagerstrom's
Tolerance
Questionnaire
(Fagerstrom
&
Schneider,
1989)
to
participate
in
the
study.
Subjects
were com-
pensated
for
their
participation
with
monetary
payment
that
was
not
contingent
upon
any
particular
type
of
performance
in
the
experi-
mental
session.
Table
1
lists
these
and
other
relevant
characteristics
of
the
subjects.
An
Apple
IIe®
microcomputer
controlled
and
obtained
data
from
a
response
console
(61
cm
by
30
cm
by
46.5
cm)
that
contained
three
Lindsley
plungers
(Gerbrands
No.
G6310;
centered
from
left
to
right
on
the
face
of
the
response
console).
Subjects
responded
on
the
left
and
right
plungers
only.
Responses
made
on
the
center
manipulandum
produced
no
pro-
grammed
effect.
Sessions
were
conducted
in
rooms
that
contained
one
response
console,
overhead
fluorescent
lighting,
a
desk
lamp,
several
current
magazines,
the
daily
local
newspaper,
and
a
radio.
Carbon
monoxide
from
expired
air
samples
from
the
lungs
was
measured
using
a
MiniCO®
Carbon
Mon-
oxide
Breathkit
(produced
by
Catalyst
Re-
search
Corporation).
Throughout
the
experiment,
each
subject
had
access
to
two
brands
of
cigarettes.
The
first
brand
was
the
subject's
own
brand.
The
second
brand
was
determined
either
prior
to
or
following
the
first
session.
The
subject
was
provided
with
three
brands;
each
was
different
than
his
or
her
own
brand
but
had
an
equiv-
alent
nicotine
rating
(based
on
Federal
Trade
Commission
[FTC]
ratings,
1991).
The
sub-
ject
was
asked
to
rate
the
three
brands
from
most
to
least
preferred.
Following
this
assess-
ment,
the
least
preferred
brand
was
provided
to
the
subject
as
the
second
brand
(referred
to
as
"other"
brand
below).
Those
subjects
who
completed
this
assessment
after
the
first
session
were
provided
with
one
of
these
three
brands
(chosen
arbitrarily)
during
the
first
session.
Procedure
Subjects
participated
in
a
minimum
of
11
3-hr
sessions
(range,
1 1
to 18),
one
session
per
day.
The
number
of
sessions
varied
due
to
subject
dropouts
(Subject
PZ)
and
differences
between
subjects
in
the
time
necessary
to
reach
stability.
Sessions
were
conducted
3
days
per
week
(either
Tuesday,
Thursday,
and
Friday
or
Monday,
Wednesday,
and
Friday).
Prior
to
each
experimental
session,
the
subject
was
required
to
abstain
from
smoking
to
reach
a
carbon
monoxide
(CO)
level
of
50%
or
lower
than
his
or
her
nonabstinence
level
(obtained
during
the
initial
interview).
This
technique
was
used
because
expired
air
CO
has
been
shown
to
correlate
with
the
number
of
ciga-
rettes
smoked
(Henningfield,
Stitzer,
&
Grif-
fiths,
1980).
Subjects
were
instructed
to
ab-
stain
from
cigarette
smoking
for 5
to
6
hr
in
order
to
meet
this
criterion.
If
a
subject
failed
to
reduce
his
or
her
CO
to
the
required
level,
the
session
was
canceled
without
payment
and
was
rescheduled.
Next,
each
subject
took
one
uniform
puff
on
his
or
her
own
cigarette
30
min
prior
to
the
start
of
the
session
(cigarettes
were
provided
by
the
experimenter).
This
30-min
presession
puff
was
used
to
equate
the
time
from
last
cigarette
exposure
across
subjects
(see
Hen-
ningfield
&
Griffiths,
1981).
485
R.
J.
DeGRANDPRE
et
al.
Instructions.
On
the
1st
day
of
the
experi-
ment
(Session
1),
subjects
were
instructed:
In
this
study
you
can
respond
on
the
left
and
right
levers
to
earn
puffs.
You
will
have
a
cer-
tain
amount
of
money
provided
to
you
to
spend
on
puffs
during
the
session.
On
different
days
the
amount
of
money
may
vary
and
so
will
the
cost
of
the
puffs
for
both
types
of
puffs.
Prior
to
beginning
each
day
you
will
be
provided
with
this
sheet,
which
will
inform
you
about
the
parameters
above.
When
you
have
earned
puffs,
the
cost
of
these
puffs
will
be
subtracted
from
the
amount
you
have
remaining,
which
will
be
shown
on
the
screen.
Also,
when
you
have
earned
puffs
you
will
have
5
minutes
to
take
them.
You
will
be
instructed
to
inhale,
hold
for
5
seconds,
and
exhale.
The
screen
will
count
down
the
300
seconds
and
will
tell
you
when
you
can
respond
again
by
displaying
"You
may
respond
now."
Also
prior
to
the
first
session
and
all
other
sessions,
subjects
were
given
the
following
daily
parameters
for
that
session:
1.
Your
brand
(left
lever/one
response
re-
quired)
Puffs
cost:
cents
for
two
puffs.
2.
Other
brand
(right
lever/one
response
required)
Puffs
cost:
cents
for
two
puffs.
3.
Today
you
have
$
to
spend
on
puffs.
Daily
procedure.
The
sequence
of
conditions
varied
across
subjects
because
the
procedure
was
modified
after
the
first
3
subjects
UH,
KS,
and
WR)
completed
the
study.
The
in-
come
conditions
in
the
modified
procedure
were
more
systematic
than
in
the
earlier
procedures
and
thus
will
be
described
first.
All
subjects
who
completed
the
modified
procedure
completed
a
minimum
of
four
in-
come
conditions:
baseline
income
(Condition
A),
one
half
baseline
income
(Condition
B),
one
third
baseline
income
(Condition
C),
and
one
sixth
baseline
income
(Condition
D).
These
percentages
of
income
were
rounded
up
to
the
nearest
$0.50.
For
each
of
these
income
con-
ditions,
price
of
the
subject's
own
and
other
puffs
was
$0.50
and
$0.10,
respectively,
for
two
puffs.
Income
for
the
initial
baseline
ses-
sion
was
set
at
$15.00.
Income
for
all
remain-
ing
baseline
sessions
(i.e.,
baseline
income)
was
that
amount
of
money
spent
in
the
first
session
rounded
up
to
the
nearest
$0.50.
The
exper-
imental
design
for
these
income-manipulation
conditions
can
be
described
as
an
ABACAD
design.
That
is,
after
establishing
a
baseline,
the
three
income
conditions
were
examined
with
baseline
conditions
interposed
between
them.
Each
baseline
condition
was
maintained
until
choice
of
the
own
puffs
was
stable
(i.e.,
plus
or
minus
one
self-administration
of
own
puffs
over
three
sessions;
JH's
data
did
not
meet
this
stability
criterion).
The
sequence
of
the
three
income
conditions
varied
across
sub-
jects
in
a
mixed
order
(see
Table
2).
Each
condition
was
completed
in
a
single
session
(except
for
baseline
sessions).
Following
the
completion
of
these
income
conditions,
some
subjects
completed
additional
conditions.
First,
additional
income
conditions
were
used
to
assess
the
effects
of
a
more
com-
plete
range
of
income
amounts
on
drug
choice
(see
Table
2).
Second,
a
single
session
was
conducted
in
which
the
price
of
the
other
puffs
was
raised
from
$0.10
to
$0.25
(for
two
puffs)
at
one
of
the
income
amounts
previously
tested
(i.e.,
Conditions
B,
C,
or
D).
The
price
of
the
other
puffs
was
increased
to
examine
whether
the
price
manipulation
would
produce
an
in-
crease
or
decrease
in
consumption
of
that
good
relative
to
choice
at
the
lower
price
at
the
same
income
amount.
The
procedure
for
the
3
subjects
who
par-
ticipated
first
was
identical
to
the
above
pro-
cedure
except
that
(a)
baseline
income
was
an
amount
that
constrained
subjects'
choice
re-
sponses
for
their
own
puffs
such
that
income
manipulations
consisted
of
increases
in
income
(instead
of
manipulations
that
decreased
in-
come),
(b)
subjects
obtained
one
puff
per
self-
administration
(instead
of
two
puffs),
(c)
the
income
conditions
were
selected
in
a
less
sys-
tematic
fashion
(i.e.,
not
necessarily
one
half,
one
third,
and
one
sixth
of
baseline
income),
and
(d)
fewer
income
manipulations
were
made
(see
Table
2).
Puffing
procedure.
After
responding
for
puffs,
the
subject
lit
one
of
the
cigarettes
he
or
she
was
provided,
but
did
not
inhale.
The
subject
was
required
to
take
one
puff
(first
3
subjects)
or
two
puffs
within
5
min
according
to
a
uni-
form-puff
procedure
in
which
the
subject
took
one
uniform
puff,
inhaled,
retained
the
smoke
in
the
lungs
for
5
s,
and
then
exhaled
(Griffiths,
Henningfield,
&
Bigelow,
1982).
When
a
sec-
ond
puff
was
allowed,
it
was
taken
25
s
after
the
subject
exhaled
the
first
puff.
Also,
the
experimenter
observed
and
recorded
the
sub-
ject's
puff
topography
(e.g.,
shallow,
normal,
486
EFFECTS
OF
INCOME
ON
CHOICE
ci
4~~~~~~~~~4.
0
4
40
4
.
.....I-
ci
"
-N
N
"
N
44
4 4
444444
X
:X
_4
_4~~~
-
-
-
-
-
-~~~~~~~~~~~~c
487
04
0c
n-
-)-
ie
-~~~-~~
-
~
--
-C1
ci'-404
~
0
.e_
L
o
~
0
\ed04
0
0
0
0
0
0
0
0
4
0
00
00
000
0000
0 0
000
000
s
-CiCio-o
o
O
.N
-O
coO
o
O0o0o
o0oO
o o o
e~-
,I
'I
'IC
C
0
'I
'I
C)
O
0
It
\
O
O)
O
'IO
0
O)
O)
O)
O
CO
O) O)
O
CO
O O O O O
0
O
ce
r
G
)
_
_
N
_
N
Ci
N
N
Ci
O
O.
O
OO
O
O
CiOoOaoO
'IC
No
N
O
CO
O t
'I
N
CO
N
N
O
L--e
e
N
N
eq
c
N
cq
-
cq
c
N
LI
Cq
* *
-4
_t
en
cn
,
t
N
N N
N
n
n
CO
O
O
CO
O O
O
O
CO
Un
O
O)
CO
O
O
O
O
C)
C)C
O
0
0 0 0 0 0
o
O
C
)
LnC
O
)
O)
O O
CO
0000000000000000000Lf00000
OOOOOOOOOO00000
000000000000000
000000000000000000
00000
000000000000000
000000000000000000
00000
000000000000000
0000000000000
0000
009LOLf
.
.
. .
.
.
.
.
. . .
. .
. .
.
.
.
.
.
.
.
. . .
.
U
xo
Co
)
o
O
C;
C_
Co
Co
C
o
C
C
C C
6 6
5
C3
C5
C
C
C
C
C5
C
6
C5
C;
_-
_--
---
___'-
_ _ _ _
_l
___
_l
l
l -
-
-
-11111-l11--
-,11I-
l
,
,
l
l
l
O O 0
)(
)C
)O00000
000C
C
)C
>0C
)C
)C
)0
o
o
c
Um
ll
UeU
nU
~U
~U
qU
qt
q
t
~U
qU
qL
~L?U
~LiU
qU
rut
qL
qL
0
U)
U(
UX
C.)
C)
UU
_
C.)
cn
5.
v
00
0
0
-.)
0
0
0-o
R.
J.
DeGRANDPRE
et
al.
N
P4~~~
n
-
e
0-1-
PC,)
9.--
,-
^o
01
_.
^4
^
^0
^^
^0
O'
;
,O
000000000000000
\0e
00000n
n
e00004c
c
f
ooo-666666666666
i-~
-4
-4
-4
--!
-~-
~-:
L
O10
00-
N
100
00010101010)
C
N
N
-s
N N
'.
N
101100
N N
N
10
000000000000000
L0
101010
LO
01100101010101)00
*n
Lt
Lt~n
Ln Ln Ln
uN
Ln
Ln
Ln
u:m
n
6 6
.6
.
.
.
.
.
.
. . .
.
000000000000000
101001000100100010010
e
.n
n
m
o
mm
m o
me
i
o
00~
P4
-
10
I
I
I
-,-I
1,1-
N
_
LOr
-
1
N
--
ce)
"
-
b
000000000000000000
Nt
'.
0
0
00
0
00
'.0
00
00
0n
0
0
0
0
N'NN
'-
N
cN-NNNN
N
1
L
O
1010
1o
000
1
0
1n
1 0
n
10
n
10
666666666666666666
0r
t
fi
O
1010010010010010010
Lti
LO;
101010
L
101010
C1-
-10
N
101010
~-
~0
~
10'.
10610
u
z4
488
4._
0
uz
U
U
uo
0.
._
0
0
-o
-
._
t:
rA
_
00
0.
0
.
0
OUC0
.
0
U
-.
0
U
0*.
U
-o
Cl)
I-
'.0-4
"0-
.4
4
4
04
,.0
cq
10
-10c
c
00 00
0 0
N
0
N
0
N
N
00C
00
L)00
c
ef
f
-
4
ce
6
6 6 6
66
0 0
00
1
0 0 0 0
10
66 66 66
0
00
0
0
0)
C)
C
10101001010
CN
C4
"
L
L
N
I
EFFECTS
OF
INCOME
ON
CHOICE
444-
v-
v-
N
N1
444
1-
..4
g--0'
4-
P
ep
N
n
C-
Ns
L
N
I4
.I
I
I
I4
.-
s
0
0
O
O
0
0
OC)0
0C
Cl
N
C
N
en
No
It
o o
o
o o o
r-
00
0
ol
-
I'l
'IC
0
N
Co
o)
Co
o
o
oC
,%
Cn
e
n
-
N
N
oq
O
cq
cq
o1
t
"o
-I
"o
NNNN
N'C,4N
0)C)0
C
)C
)C
)0
000
O
.
.
.
.
.
-~.
..
. . .
.
00000o00000000000
00000OOOD
00
00000
C
L
n
n
Lr
C0
C
O
O
4
Cliclicl
ci
L
~6
1
\
e6
N'.ON6
-l
14
_e
N
-_
N
_
t
ULf
N
oo^_
444
4
4
N~~~~r
't
It
C)
0 0
c14
Lr
I'd
Lf
u)
-4
_r
u)
un
en
en
444
CD4CDO
-C
-C
t)
-O
UN
cl
Ln
Ln
IRt
Ln
0 0 0
~~~".
v
-
v-
--
" "
- N
o
- -
o
Lt
Lf
L(
Lo
Lo
O
M
o
n
0'
'.
LO
LO
Lf
*n
Ln
Lf
'. '.
'n
en
*n
o
LO
LO
L(
L(
Lf
Lf
L
LO
f
LO LO
Lf
O
LO
-o
o6
6 6
o6
6 6
o6
6 6 6 6
.
.
OO0000000000000o
t t
N
.
(-)
N
U6
H3
Ut
0
0
3.
U
0
0
.0
0
.
O
UU,
U
0.0
Uw
=
_.40
._
CQ2
489
U
.0
O0
._
to
0
0-
-
U
3°
U
0D
-c3o
CZ
CZ
0
Y
U.o
3U,
00
uat
0)*
U,.-w
Q
.0.
cn
cY
'!.
R.
J.
DeGRANDPRE
et
al.
10
Jo
BT
0
0
.
.
0
.
*
-
1r
*
10
JH
.
o
0
o-
U
0
0
100
.
lo*
KC
0
O
0
No0
0
100
1
r
100
1
100
.
10
Jun0
0
1
*
10
50
PZ
0
0
1
*
10
100
INCO
1
)ME
(IN
$)
Fig.
1.
Consumption
(puffs)
of
own
and
other
puffs
is
plotted
as
a
function
of
income
amount
for
all
7
subjects.
Data
at
the
baseline
income
amount
(shown
in
Table
2)
for
each
subject
represent
mean
consumption
across
all
baseline
sessions.
An
asterisk
is
shown
on
the
x
axis
to
represent
the
baseline
income.
490
100
10*
1
100
10
own
o
other
100
10
.
C,)
LI.
lLJ
U.
EL
1
100
100
JR
0
U
0
*
10
KS
100
1
0
0
.
*
10
WR
100
10
1I
1
-1
100
10
100
U
0
0
*10
100
EFFECTS
OF
INCOME
ON
CHOICE
or
deep
puffs)
during
the
first
session
and
in-
structed
the
subject
not
to
change
that
topog-
raphy.
Topography
was
monitored
through-
out
the
study,
and
the
subject
was
told
of
deviations.
RESULTS
Effects
of
Income
When
choice
of
the
two
reinforcers
(own
puffs
and
other
puffs)
is
plotted
as
a
function
of
income
amount
(see
Figure
1),
the
magni-
tude
of
the
income
effect
is
an
orderly
function
of
this
variable
(data
are
plotted
in
log-log
coordinates
to
show
proportional
change
such
that
the
slope
of
the
line,
in
economic
terms,
equals
its
elasticity
coefficient;
see
Samuelson
&
Nordhaus,
1985).
In
all
but
two
cases,
every
successive
increase
in
income,
from
the
lowest
income
amount
to
the
highest
income
amount,
increased
choice
responses
for
the
own
puffs.
In
those
cases
in
which
an
increase
did
not
occur
(low-income
conditions
for
BT
and
KC),
no
responses
were
made
for
the
own
puffs
across
the
two
consecutive
income
amounts
(i.e.,
floor
effect).
A
similar
function
exists
in
the
opposite
direction
for
the other
puffs,
albeit
more
variable
(i.e.,
increases
in
income
de-
creased
choice
responses
for
the
other
puffs).
Several
subjects'
data
for
both
own
puffs
and
other
puffs
suggest
an
asymptote
in
consump-
tion
at
high
and
low
income
amounts,
respec-
tively
(e.g.,
BT
and
KC).
To
illustrate
the
across-subject
similarity
of
this
function
for
the
own-puff
data,
these
data
for
all
7
subjects
shown
in
Figure
1
are
collapsed
into
a
single
function
in
Figure
2.
The
effects
of
income
on
consumption
of
subjects'
own
puffs
are
mark-
edly
similar
across
subjects.
In
order
to
illustrate
better
the
control
ex-
erted
by
the
income
manipulation
over
choice
responding
across
sessions
for
each
subject,
the
data
plotted
in
Figure
1
are
plotted
again
in
Figures
3
and
4
as
a
function
of
session
order.
Total
puffs
(own
puffs
+
other
puffs)
are
also
plotted
for
each
subject.
Recall
that
Subjects
JR,
KS,
and
WR
participated
in
the
original
procedure.
Changes
in
income
across
sessions
signifi-
cantly
altered
choice
responding.
Again,
de-
creases
in
income
from
the
baseline
amount
decreased
choice
responses
for
their
own
puffs
z
0
IL.
0
z
0
(0
z
0
100
10
1.
.
U
*
U
1
U
o
;
U
3B
U'
*
:
Em
10
INCOME
(IN
$)
*
JH
i
JR
*
KC
o
KS
LPZ
o
WR
100
Fig.
2.
Consumption
(puffs)
of
own
puffs
for
all
7
subjects
is
plotted
together
as
a
function
of
income
amount.
These
data
are
also
shown
in
Figure
1
separately
for
each
subject.
Data
at
the
baseline
income
amount
(shown
in
Table
2)
for
each
subject
represent
mean
consumption
across
all
baseline
sessions.
and
increased
choice
responses
for
the
other
puffs.
In
many
cases,
when
the
session
income
was
radically
different
from
the
previous
ses-
sion's
or
the
following
session's
income,
the
reinforcer
chosen
reversed,
such
that
the
re-
inforcer
preferred
in
one
session
was
opposite
to
that
of
the
next
session.
The
effects
of
income
on
total
consumption
were
varied.
First,
the
income
manipulations
had
little
effect
on
2
subjects'
(BT
and
WR)
total
consumption,
even
though
choice
was
sig-
nificantly
affected.
Note
that
Subject
BT's
to-
tal
consumption
was
affected
at
very
low
and
very
high
incomes.
Also,
Subject
WR's
base-
line
levels
of
consumption
shifted
in
the
middle
of
the
experiment
and
returned
to
his
original
levels
toward
the
end
of
the
study.
Second,
3
subjects'
(JH,
KC,
and
PZ)
total
consumption
generally
decreased
as
income
decreased.
However,
total
consumption
for
2
of
these
sub-
jects
(KC
and
PZ)
was
generally
constant
ex-
cept
when
income
was
so
low
that
baseline
levels
of
consumption
were
not
mathematically
possible.
Subject
JH,
whose
choice
responding
did
not
meet
stability
requirements
after
the
first
income
manipulation
(one
third
condi-
tion),
showed
gradual
decreases
in
consump-
tion
as
the
experiment
progressed.
Finally,
for
Subjects
JR
and
KS,
the
variability
in
total
puffs
produced
by
the
income
manipulations
is
similar
in
magnitude
to
the
variability
in
total
puffs
across
the
baseline
conditions.
This
491
R.
J.
DeGRANDPRE
et
al.
60
36
50
151
JH
50
40
1
40
-2
1s
1
12
3
30'
6
12
30
6
4
15
~~~~~~~
2~~~
60
-
20
10
-
10-
0
_O-
puffs
60.'II
60'I
'
III
50
0
40-
40
30
3
,
,
,
, ,
,
,
,
0
2
4
681010
12
14 16
18
20
0
2
4
6
8
10
12
14
16
18
20
30
KCS16.5
3
PZ
6
6
15
555.5
5.5
15
6
6
6
55
5.5
5.5
5.5
5.5
5.56
20
-
1
23
20
5
~~~~~~~~~~~~2
10'
10'
50.5
own
puffs
40
40
-o--
other
puffs
~.total
puffs
30
30'
20
-20
02
4 6
8
1014214
1680024681
246
18200
SESSION
492
EFFECTS
OF
INCOME
ON
CHOICE
JR
2.5
2.5
2.5
2.5
2.5
2
2.5
2
22.5
2.52.
25
2.5
2.5
2.5
10-
2
4
6
8
10
12
14
16
18
20
(
M
2
4
6
8
10
12
*
own
puffs
o
othr
puffs
total
puffs
0
2
4
6
8
10
12 14
16
18
20
SESSION
Fig.
4.
Consumption
(puffs)
of
the
two
reinforcers
(own
and
other
puffs)
is
plotted
as
a
function
of
session
for
the
3
subjects
who
participated
under
the
original
procedure
(top
panels).
The
income
amount
for
each
session
is
indicated
in
dollar
amounts
just
above
one
data
point
for
each
session.
Also,
total
puffs
(own
+
other
puffs)
are
plotted
as
a
function
of
session
(lower
panel).
The
data
for
sessions
in
which
price
was
varied
are
not
shown.
variability
makes
it
difficult
to
interpret
the
effects
of
income
on
total
consumption.
Another
aspect
of
the
total-consumption
data
is
the
finding
that
subjects
sometimes
did
not
spend
their
entire
income
even
in
sessions
in
which
they
responded
for
both
types
of
puffs
(see
"$
not
spent"
and
"distribution
within
session"
in
Table
2).
This
result
appeared
to
occur
primarily
because
subjects
conserved
their
income
to
ensure
that
they
could
obtain
puffs
throughout
the
session.
As
a
result
of
this
tac-
tic,
subjects
frequently
ran
out
of
time
in
the
session
prior
to
spending
their
allotted
income.
For
example,
even
when
subjects
were
allotted
an
income
that
was
almost
(or
more
than)
enough
to
respond
exclusively
for
the
more
expensive
puffs
(i.e.,
own
puffs),
they
still
of-
ten
purchased
some
or
many
of
the
cheaper
brand
(i.e.,
other
puffs),
even
though
their
overall
data
showed
preferences
for
the
more
expensive
brand
(i.e.,
increases
in
own
puff
consumption
with
increases
in
income).
This
occurred
especially
when
the
first
session
had
a
$15.00
income
(see
data
from
Subjects
BT,
JH,
KC,
and
PZ).
The
most
striking
example
of
this
pattern
of
responding
is
shown
by
Sub-
ject
BT's
data;
he
responded
exclusively
for
own
puffs
when
income
was
very
high.
This
tactic
is
also
indicated
by
the
distribution
of
responding
by
many
subjects
that
alternated
between
the
more
expensive
puffs
(own
puffs)
and
the
less
expensive
puffs
(other
puffs),
es-
pecially
toward
the
end
of
the
session
(see
Ta-
ble
2).
Effect
of
Price
Table
3
shows
the
effects
of
repeating
an
income
condition
when
the
price
of
the
other
20-
10-
LL-
i
]20
10-
0
0
O
-
30-
/vJV
20-
10-
Fig.
3.
Consumption
(puffs)
of
own
and
other
puffs
is
plotted
as
a
function
of
session
for
the
4
subjects
who
participated
under
the
modified
procedure
(top
panels).
The
income
amount
for
each
session
is
indicated
in
dollar
amounts
just
above
one
data
point
for
each
session.
Also,
total
puffs
(own
+
other
puffs)
are
plotted
as
a
function
of
session
(lower
panel).
The
data
for
sessions
in
which
price
was
varied
are
not
shown.
493
R.
J.
DeGRANDPRE
et
al.
Table
3
Results
of
price
manipulation.
Relative
Quantity
quantity
consumed
consumeda
Low
High
Low
High
Subject
Brand
price
price
price
price
BT
own
20b
12
54b
33
other
17b
24
46b
67
JH
own
12
12
37.5
60
other
20
8
62.5
40
KC
own
14
12
58.3
60
other
10
8
41.7
40
KS
own
3b
1
37.6b
10
other
8.13b
9
62.4b
90
PZ
own
6
6
37.5
60
other
10
4
62.5
40
WR
own
5.67b
3
30.2b
23
other
11.44b
10
69.8b
77
a
Percentage
of
total
puffs.
bMean,
based
on
multiple
sessions.
puffs
was
increased
from
$0.10
to
$0.25
(ex-
cept
Subject
KS,
whose
other-puffs
price
in-
creased
to
$0.20;
Subject
JR
did
not
complete
the
price
condition).
The
effect
of
the
price
increase
is
shown
in
absolute
terms
(con-
sumption
of
own
and
other
puffs)
and
in
rel-
ative
terms
(percentage
of
total
consumption
for
own
and
other
puffs).
If
multiple
sessions
of
the
income
condition
were
completed
at
the
low
price
(i.e.,
if
baseline
income
was
used),
the
data
for
the
low
price
represent
a
mean.
Increasing
the
price
of
the
other
puffs
in-
creased
the
absolute
consumption
of
the
other
puffs
by
Subjects
BT
and
KS
and
relative
con-
sumption
by
Subjects
BT,
KS,
and
WR.
For
example,
BT
increased
consumption
of
the
other
puffs
from
a
mean
of
17
puffs
to
24
puffs;
WR
increased
relative
consumption
of
the
other
puffs
from
69.8%
to
77%.
Importantly,
the
absolute
or
relative
consumption
at
the
high
price
does
not
appear
to
be
related
simply
to
the
absolute
or
relative
consumption
at
the
low
price.
That
is,
although
Subjects
KS
and
WR-
who
showed
relative
increases
in
consumption
of
the
other
puffs
at
the
higher
price-chose
more
of
the
other
puffs
at
the
low
price,
so
did
2
of
the
3
subjects
(JH
and
PZ)
who
did
not
show
relative
or
absolute
increases
(see
Table
3).
To
examine
these
individual
differences,
the
acceptability
of
the
other
puffs
was
assessed
using
postsession
ratings
of
"how
much
they
liked"
their
own
puffs
and
the
other
puffs.
This
was
done
by
asking
subjects
to
respond
to
this
question
on
two
100-mm
visual
ana-
logue
scales
(VAS;
100
=
"very
much")
that
corresponded
to
the
two
brands
of
puffs.
These
data
were
collected
only
for
the
4
subjects
who
participated
under
the
modified
procedure.
The
mean
"liking"
scores
across
all
sessions
for
the
own
and
other
puffs,
respectively,
were
81.2/
57.5
for
Subject
BT
(who
showed
an
increase
in
consumption
of
the
other
puffs),
82.8/0.0
for
Subject
JH,
89.9/1.83
for
Subject
KC,
and
87.6/0.88
for
Subject
PZ
(all
of
whom
showed
decreased
consumption
of
the
other
puffs).
When
liking
scores
for
the
own
or
other
puffs
are
removed
for
those
sessions
in
which
own
or
other
puffs
were
not
chosen,
these
results
do
not
change.
That
is,
subjects
reported
very
similar
liking
scores
regardless
of
whether
they
smoked
a
particular
brand
during
the
session.
Finally,
to
examine
the
possibility
that
an
increase
in
price
may
be
functionally
equiva-
lent
to
a
decrease
in
income,
a
post-hoc
analysis
was
done
that
compared
consumption
during
the
high-price
condition
(shown
in
Table
2)
to
the
consumption
predicted
from
subjects'
income
functions
(shown
in
Figure
1).
To
do
this,
the
increase
in
price
was
transformed
into
a
decrease
in
income
by
using
the
observed
consumption
of
both
brands
at
the
higher
price
to
determine
the
income
that
would
have
been
spent
in
the
price
condition
if
the
other
good
was
at
its
previous
price
of
$0.
10
per
two
puffs.
For
example,
BT
spent
$3.00
on
both
brands
during
the
price
condition
at
an
income
of
$6.00.
Thus,
if
the
other
puffs
had
cost
their
normal
price
($0.10
per
two
puffs),
BT
would
have
required
an
income
of
only
$4.20
to
pro-
duce
this
result.
Using
this
amount
($4.20),
if
the
price
and
income
manipulation
were
func-
tionally
equivalent
for
this
subject,
consump-
tion
at
$6.00
income
(at
the
high
price
for
other
puffs)
should
equal
consumption
at
$4.20
at
the
low
price.
To
test
this,
we
can
plot
BT's
data
from
the
price
condition
at
an
income
of
$4.20
along
with
the
income
data
shown
in
Figure
1.
For
the
3
subjects
who
showed
relative
in-
creases
in
consumption
for
the
other
puffs
dur-
ing
the
price
condition
(BT,
KS,
and
WR),
consumption
was
consistent
with
the
respective
income
functions
for
own
and
other
puffs
(see
494
EFFECTS
OF
INCOME
ON
CHOICE
*
own
(low
price)
O
other
(low
price)
M
own
(high
price)
Q
other
(hlgh
price)
BT
.
o
U
U
*
-
100
.
10
0
0
10
KC
I
100
100
O
0i-
0
00
m~~
1
10
50
PZ
0
00
0
U
-
10
100
10
1
100
10
JH
.
O
O
!0
-
o0
-
1
0
0
10
KS
100
Ooo
U
s
.
1
10
WR
U
0
1
10
100
100
INCOME
(IN
$)
Fig.
5.
The
same
data
plotted
in
Figure
1
are
shown
for
the
subjects
who
completed
the
price
condition
(small
symbols)
along
with
the
data
from
the
price
manipulation
(large
symbols).
The
latter
data
are
plotted
at
the
income
amount
that
would
have
been
spent
if
puffs
were
only
$0.10.
Own-puff
and
other-puff
data
at
the
baseline
income
amounts
are
means
across
all
baseline
sessions.
Figure
5).
However,
data
from
the
3
subjects
who
showed
decreases
in
relative
consumption
of
the
other
puffs
UH,
KC,
and
PZ)
were
not
consistent
with
their
respective
income
func-
tions
for
the
own
and
other
puffs.
Instead,
all
3
subjects'
consumption
of
own
puffs
was
greater
than
predicted
by
their
income
function
for
those
puffs,
whereas
these
subjects'
con-
sumption
of
other
puffs
was
clearly
lower
than
predicted
by
the
function
for
the
other
puffs.
1
100'
10
I
100
n
U-
=
10-
1'-
100
-
10
0
1
495
R.
J.
DeGRANDPRE
et
al.
DISCUSSION
The
main
results
can
be
summarized
as
fol-
lows:
First,
as
income
increased,
consumption
of
own
puffs
increased
while
consumption
of
the
less
expensive
other
puffs
decreased.
The
magnitude
of
these
within-subject
effects
was
an
orderly
function
of
income.
For
some
sub-
jects,
these
effects
of
income
on
choice
had
little
effect
on
total
drug
consumption.
Second,
in-
creases
in
the
price
of
the
other
puffs
resulted
in
a
decrease
of
other
puff
consumption
in
4
subjects,
but
resulted
in
an
increase
in
other
puff
consumption
in
the
other
2
subjects
who
completed
this
condition.
As
outlined
in
the
introduction,
the
effects
of
income
and
price
manipulations
have
been
conceptualized
in
economic
terms
in
previous
studies
(e.g.,
Silberberg
et
al.,
1987).
Goods
whose
consumption
increases
as
income
in-
creases
are
typically
defined
as
normal
goods
(i.e.,
goods
for
which
income
and
consumption
are
directly
related),
whereas
goods
whose
con-
sumption
decreases
as
income
increases
are
typically
defined
as
inferior
goods
(i.e.,
goods
for
which
income
and
consumption
are
in-
versely
related).
Consequently,
own
puffs
in
the
present
study
can
be
defined
as
a
normal
good
and
other
puffs
as
an
inferior
good
for
all
7
subjects.
As
shown
in
Figures
1
and
2,
the
effects
of
income
on
choice
can
be
plotted
so
as
to
depict
the
nature
of
these
goods.
These
functions,
in
which
consumption
is
plotted
as
a
function
of
income,
are
defined
in
economic
terms
as
Engel
curves.
Moreover,
the
effect
of
increasing
consumption
of
an
inferior
good
by
increasing
its
price
is
termed
a
Giffen-good
ef-
fect.
That
is,
when
the
slope
of
a
demand
(price)
function
is
positive-as
opposed
to
the
negative
slope
predicted
by
the
law
of
demand-the
good
is
acting
as
a
Giffen
good
(Lea
et
al.,
1987;
cf.
Gilley
&
Karels,
1991).
Such
an
effect
was
shown
here
in
2
of
6
subjects.
Note
that
when
the
slope
of
a
demand
(price)
function
is
negative,
the
good
is
acting
as
an
ordinary
good.
The
implications
of
these
effects
and
conceptualizations
are
discussed
below.
Income
Manipulations
The
income
manipulation
employed
in
pre-
vious
research
on
the
effects
of
income
on
choice
has
been
trials
per
session
(Hastjarjo
&
Sil-
berberg,
1992;
Hastjarjo
et
al.,
1990;
Silber-
berg
et
al.,
1987).
The
income
manipulation
in
this
study
differed
in
that
a
medium-of-
exchange
procedure
was
used
in
which
sub-
jects
were
provided
with
some
income
to
spend
throughout
the
session
via
a
discrete
response
for
one
or
both
of
two
differently
priced
re-
inforcers.
Despite
these
differences,
at
least
four
other
studies
have
demonstrated
normal
and
inferior
goods
using
a
concurrent-operant
arrangement
(Battalio
et
al.,
1991;
Hastjarjo
&
Silberberg,
1992;
Hastjarjo
et
al.,
1990;
Sil-
berberg
et
al.,
1987).
By
demonstrating
similar
effects
with
a
new
species
(humans)
and
a
new
and
important
class
of
reinforcers
(drug),
this
study
further
demonstrates
the
utility
of
the
income
concep-
tualization
in
the
study
of
choice.
Concerning
species
generality,
replicating
the
effects
of
in-
come
on
choice
with
humans
is
of
obvious
im-
portance.
The
present
study
not
only
shows
the
effects
of
income
previously
shown
in
an-
imals
but
goes
further
in
demonstrating
more
complete
functional
relations
between
income,
price,
and
choice
using
a
medium-of-exchange
procedure
common
to
everyday
human
envi-
ronments.
Concerning
reinforcer
generality,
the
finding
that
drug
reinforcers
are
affected
by
income
in
much
the
same
way
as
food
re-
inforcers
are
indicates
that
these
effects
of
in-
come
are
general
and
appear
to
be
applicable
to
most
or
all
primary
reinforcers.
Finally,
the
different
effects
of
income
on
concurrently
available
reinforcers
were
well
described
by
income
functions
(i.e.,
Engel
curves)
and
the
characterization
of
inferior
or
normal
goods.
This
conceptualization
is
useful
in
that
(a)
it
provides
a
framework
to
organize
the
differ-
ential
effects
of
income
across
a
range
of
in-
come
magnitudes
that
produce
complete
para-
metric
functions,
and
(b)
the
two
functions
for
normal
and
inferior
goods
provide
an
exper-
imental-economic
model
or
baseline
for
fur-
ther
complex
analyses,
such
as
the
interaction
between
income
and
other
more
traditional
factors
shown
to
influence
choice
(discussed
below).
The
functions
for
normal
and
inferior
goods
provide
an
important
demonstration
that
a
re-
inforcer
is
not
endowed
with
inherent
prop-
erties
independent
of
the
historical
and
current
context
in
which
it
controls
behavior.
For
ex-
ample,
in
the
present
study,
income
reversed
preference
between
the
two
reinforcers
such
that
reinforcer
effectiveness,
conventionally
defined
(see
Katz,
1990),
depended
on
income
and
not
on
any
aspects
of
the
reinforcers
per
496
EFFECTS
OF
INCOME
ON
CHOICE
se
(e.g.,
within-session
rate
or
magnitude
of
reinforcement).
Given
that
the
availability
of
a
drug
is
almost
always
constrained
in
drug
self-administration
studies,
the
choice
observed
between
the
concurrently
available
reinforcers
may
be
due,
in
part,
to
the
prevailing
income.
This
is
not
to
say,
however,
that
qualitative
and
quantitative
aspects
of
the
reinforcer
and
reinforcement
schedule
are
not
important.
Rather,
the
effect
a
reinforcer
has
on
respond-
ing
is
determined
by
a
confluence
of
these
en-
vironmental
variables.
This
notion
of
intrinsic
properties
also
ap-
plies
to
the
behavioral-economic
terms
of
in-
ferior,
normal,
and
Giffen
goods.
Although
useful
in
categorizing
reinforcers
relative
to
one
another
across
different
situations,
these
terms
describe
a
relationship
among
choice,
price,
and
income
with
respect
to
a
particular
context
and
thus
do
not
describe
inherent
prop-
erties
of
the
reinforcer
per
se
(cf.
Hastjarjo
et
al.,
1990).
For
example,
although
the
other
puffs
were
an
inferior
good
for
all
7
subjects
in
the
present
study,
this
good
would
likely
be
a
normal
good
in
a
concurrent
arrangement
in
which
these
puffs
were
available
along
with
puffs
with
no
nicotine
content.
The
effects
of
income
on
choice
also
have
relevance
to
issues
concerning
the
above
lia-
bility
of
drugs.
Abuse
liability
refers
to
the
like-
lihood
that
any
given
pharmacological
agent
will
be
self-administered
by
humans
and
will
be
used
in
a
nonprescribed
fashion
for
nonmedical
purposes.
The
drug
self-admin-
istration
preparation
is
one
method
for
ascer-
taining
abuse
liability,
based
on
the
finding
in
behavioral
pharmacology
that
drugs
of
abuse
can
serve
as
positive
reinforcers
and
are
self-
administered
by
human
and
nonhuman
ani-
mals
in
laboratory
settings
(Griffiths,
Bigelow,
&
Henningfield, 1980;
Young
&
Herling,
1986).
The
present
data
are
relevant
to
this
issue
because
they
demonstrate
important
ef-
fects
of
a
variable
not
widely
shown
to
affect
drug
consumption
in
laboratory
settings
(viz.
income).
By
showing
that
decreases
in
income
in
some
contexts
may
actually
increase
choice
of
an
inferior
good
relative
to
a
normal
good,
these
data
suggest
that
laboratory
analyses
of
abuse
liability
may
need
to
examine
liability
in
a
variety
of
choice
contexts.
For
example,
at
high
income
levels
in
the
present
study,
the
other
puffs
appeared
to
have
low
abuse
lia-
bility
relative
to
the
subjects'
own
puffs;
how-
ever,
when
income
decreased,
preference
re-
versed,
thus
altering
the
assessment
of
abuse
liability.
In
terms
of
drug
taking
and
drug
policy,
these
findings
suggest
that
decreases
in
the
drug
user's
income
(e.g.,
unemployment)
may
not
affect
drug
intake
(i.e.,
decrease
it)
but
rather
may
affect
drug
choice
(e.g.,
switching
from
cocaine
to
crack
cocaine).
This
finding
also
suggests
that
the
development
and
production
of
lower
priced
drug
substitutes
(e.g.,
crack
cocaine)
may
increase
as
a
result
of
increases
in
price
of
the
already
available
drug,
decreases
in
the
supply
of
the
latter,
or
decreases
in
income.
Price
Manipulations
The
analysis
of
the
effects
of
price
in
the
present
experiment
is
very
preliminary,
yet
it
is
an
initial
step
toward
addressing
the
inter-
action
between
price
and
income
factors
in
determining
drug
consumption
and
the
con-
ditions
under
which
price
and
income
manip-
ulations
are
functionally
equivalent.
The
Gif-
fen-good
effect
shown
in
2
subjects
has
been
shown
in
previous
experimental
analyses.
For
example,
in
Experiment
2
of
Silberberg
et
al.
(1987),
price
was
manipulated
by
varying
the
probability
of
reinforcement
for
responding
on
the
lever
producing
the
bitter
pellet.
By
in-
creasing
the
price
of
the
bitter
pellet,
choices
for
the
bitter
pellet
increased,
thus
demon-
strating
a
Giffen-good
effect.
Similar
effects
were
also
shown
in
rats
by
Hastjarjo
et
al.
(1990)
and
Battalio
et
al.
(1991).
As
already
stated,
Giffen-good
effects
are
paradoxical
in
that
they
are
defined
by
posi-
tively
sloping
demand
(price)
curves,
opposite
in
direction
to
those
predicted
by
the
law
of
demand
(Allison,
1983;
Lea
et
al.,
1987).
That
increases
in
price
of
a
reinforcer
can
increase
its
consumption
is
also
perplexing
in
behavior-
analytic
terms,
because
decreases
in
rate
of
reinforcement
are
assumed
to
decrease,
not
in-
crease,
response
rate
or
strength
(Battalio
et
al.,
1991;
Williams,
1988).
Despite
these
ap-
parent
paradoxes,
however,
the
present
study
suggests
that
the
Giffen-good
effect
is
quan-
tifiable
and
that
an
experimental
analysis
of
such
dynamic
relations
between
price
and
in-
come
is
possible.
For
example,
the
price
ma-
nipulation
was
shown
here
to
be
functionally
equivalent
to
an
income
manipulation
for
3
subjects
(2
of
which
showed
Giffen-good
ef-
497
R.
J.
DeGRANDPRE
et
al.
fects).
This
conclusion
is
based
on
two
find-
ings.
First,
when
price
of
the
other
puffs
in-
creased,
Subjects
BT,
KS,
and
WR
increased
the
proportion
of
money
spent
on
other
puffs
(46%
-
67%,
62.4%
-
90%,
69.8%
-
77%),
whereas
Subjects
JH,
KC,
and
PZ
decreased
the
proportion
spent
(62.5%
-
40%,
41.7%
-
40%,
62.5%
-
40%).
Interestingly,
all
3
of
the
latter
subjects
had
a
60%
-
40%
distribution
between
own
and
other
puffs
at
the
high
price
(see
Table
3).
Second,
the
data
from
the
price
condition
were
consistent
with
the
income
functions
for
the
3
subjects
who
showed
an
increase
in
their
relative
consumption
of
the
other
puffs.
The
other
3
subjects'
outcomes,
however,
were
in-
consistent
with
their
respective
income
func-
tions.
One
factor
that
was
associated
with
whether
subjects
showed
functional
equiva-
lence
(or
the
Giffen-good
effect)
was
their
self-
reported
acceptability
of
the
other
puffs.
The
1
subject
(BT)
who
completed
the
VAS
scales
and
showed
functional
equivalence
reported
that
the
other
good
was
more
acceptable
than
the
3
remaining
subjects
who
did
not
show
functional
equivalence
or
Giffen-good
effects.
Below
is
an
account
of
why
subjects'
relative
or
absolute
preferences
might
increase
as
the
price
of
that
reinforcer increases.
Economic
theory
suggests
that
Giffen-good
effects
are
most
likely
to
occur
when
a
signif-
icant
proportion
of
one's
income
is
allocated
to
an
essential
but
inferior
good
(Lea
et
al.,
1987).
A
Giffen
good
is
observed
when,
as
the
price
of
the
inferior
good
increases,
the
con-
sumer
compensates
for
the
reduced
purchasing
power
by
reducing
consumption
of
a
more
ex-
pensive
normal
good
in
favor
of
purchases
of
the
(still
cheaper)
inferior
good;
this
substi-
tution
might
occur
because
the
normal
good
and
the
inferior
good
share
some
of
the
same
essential
features.
Consider,
for
example,
Sub-
ject
BT
in
the
present
study.
This
subject
con-
sumed
about
20
puffs
of
both
the
own
($0.50
for
two
puffs)
and
other
puffs
($0.10
for
two
puffs)
at
the
$6.00
income.
Thus,
he
spent
about
$5.00
on
own
puffs
and
$1.00
on
other
puffs.
Now,
if
the
price
of
the
other
puffs
is
raised
to
$0.25
at
this
income
amount,
one
can
mathematically
determine
choice
between
the
two
brands
if
he
is
to
maintain
a
total
of
40
puffs
while
simultaneously
maximizing
con-
sumption
of
the
normal
good.
According
to
this
analysis,
BT
should
decrease
consumption
of
own
puffs
to
eight
puffs
at
a
cost
of
$2.00
(8/2
x
$0.50
=
$2.00)
and
increase
consumption
of
other
puffs
to
32
puffs
at
a
cost
of
$4.00
(32/2
x
$0.25
=
$4.00).
In
actuality,
BT
did
decrease
own
puffs
(to
12
puffs)
and
increased
consumption
of
other
puffs
(to
24
puffs).
Although
Subject
BT
showed
a
clear
Gif-
fen-good
effect,
the
2
other
subjects
(KS
and
WR)
who
showed
functional
equivalence
be-
tween
income
and
price
did
not
show
as
robust
an
effect
of
price
on
choice.
One
interpretation
of
this
finding
is
that
Giffen-good
effects
rep-
resent
one
end
of
a
continuum
in
which
sub-
jects
respond
in
a
manner
that
maintains
a
constant
overall
intake
of
something
provided
by
both
reinforcers
(e.g.,
nicotine).
At
the
other
end
of
this
continuum,
subjects
respond
in
a
manner
that
maintains
a
constant
intake
of
the
normal
good.
According
to
this
continuum,
the
6
subjects
who
completed
this
condition
could
be
ranked
according
to
relative
consumption
of
the
other
puffs
across
the
low
and
high
price
as
follows:
KS,
BT,
WR,
KC,
JH,
and
PZ
(see
Table
3).
This
discussion
highlights
the
complexity
of
the
interaction
of
price
and
income.
Clearly,
however,
more
laboratory
research
on
the
ef-
fects
of
price
might
help
to
better
determine
the
conditions
under
which
Giffen-good
effects
occur.
Such
research
might,
for
example,
ex-
amine
whether
there
is
a
window
of
prices
and
incomes
in
which
a
Giffen-good
effect
would
occur
and
whether
that
window
is
correlated
with
other
phenomena
(e.g.,
price
at
which
choice
between
the
two
reinforcers
is
at
equi-
librium).
Economic
Factors
in
the
Study
of
Choice
The
income
manipulation
in
this
and
pre-
vious
studies
is
interesting
because
the
manip-
ulation
does
not
alter
the
rate,
magnitude,
or
delay
of
reinforcement
of
either
of
the
con-
currently
available
reinforcers.
This
is
similar
to
the
procedure
of
open
and
closed
economies,
in
that
the
variable
of
openness
does
not
di-
rectly
alter
any
quantitative
or
qualitative
as-
pects
of
the
reinforcers
(Hursh
&
Bauman,
1987).
Despite
this
fact,
income
reversed
con-
sumption
in
the
present
study
such
that
the
reinforcer
that
would
be
considered
to
have
the
greatest
effectiveness
depended
on
income
and
not
on
any
aspects
of
the
reinforcers
per
se.
This
finding
thus
suggests
that
the
degree
to
which
a
stimulus
can
function
as
a
reinforcer
498
EFFECTS
OF
INCOME
ON
CHOICE
499
cannot
be
completely
predicted
by
considering
only
these
traditional
characteristics
of
rein-
forcers
(e.g.,
rate,
magnitude,
delay).
Moreover,
the
finding
that
a
price
manip-
ulation
can
have
very
different
effects
on
choice
across
subjects
(i.e.,
increase
or
decrease
pro-
portion
spent
on
the
other
puffs)
is
another
example
of
the
need
for
a
more
complete
un-
derstanding
of
how
historical
and
current
en-
vironmental
variables
interact
to
determine
choice
behavior.
Importantly,
the
effects
of
in-
come
and
price
have been
suggested
by
some
(e.g.,
Battalio
et
al.,
1991;
Silberberg
et
al.,
1987)
to
be
incongruent
with
matching
theory
(Herrnstein,
1961,
1970;
Herrnstein
&
Vaughan,
1980).
This
argument
is
based
on
the
notion
that
increases
in
price
decrease
rate
of
reinforcement
and
thus
should
decrease,
not
increase,
responding
for
the
inferior
good
(i.e.,
decreases
in
rate
of
reinforcement,
according
to
matching,
should
decrease
rate
of
respond-
ing
for
that
reinforcer).
In
contrast,
however,
it
has
been
suggested
that
the
behavioral-eco-
nomic
analyses
examine
response
distribution
(e.g.,
matching)
when
behavior
is
not
at
equi-
librium.
Although
food
deprivation
may
affect
consumption
between
imperfect
substitutes
such
as
food
and
water,
for
example,
matching
theory
is
an
account
of
the
effects
of
rate
of
reinforcement
on
organisms'
distribution
of
re-
sponding
when
the
effects
of
such
variables
(e.g.,
imperfect
substitutability,
income)
are
not
operating
(Herrnstein
&
Prelec,
19921;
Herrnstein
&
Vaughan,
1980).
Although
this
latter
point
appears
to
be
a
valid
defense
of
matching
theory,
the
effects
of
income
in
this
and
previous
behavioral-economic
analyses
suggest
that
traditional
behavioral
theories
of
choice
are
incomplete
and
should
include
a
variable
such
as
income
in
order
to
take
into
account
more
complex
environments
in
which
choice
occurs.
Variables
such
as
income
are
likely
to
operate
to
some
degree
in
most
studies,
even
though
the
possibility
of
such
influences
is
typically
ignored.
Conclusion
Applying
the
income
notion
from
economics
to
the
study
of
choice
has
provided
a
method
for
categorizing
(inferior,
normal,
ordinary,
Giffen
goods),
quantifying
(Engel
curves),
and
predicting
(functional
equivalence
between
price
and
income)
the
effects
of
a
complex
array
of
manipulations.
Future
research
will
need
to
assess
the
generality
of
these
claims,
especially
those
related
to
the
functional
equiv-
alence
between
price
and
income.
Moreover,
further
empirical
and
conceptual
work
will
be
necessary
to
assess
the
significance
of
these
data
in
generating
a
more
comprehensive
the-
ory
of
choice.
REFERENCES
Allison,
J.
(1983).
Behavioral
substitutes
and
comple-
ments.
In
R.
L.
Mellgren
(Ed.),
Animal
cognition
and
behavior
(pp.
1-30).
Amsterdam:
North-Holland.
Battalio,
R.
C.,
Kagel,
J.
H.,
&
Kogut,
C.
A.
(1991).
Experimental
confirmation
of
the
existence
of
a
Giffen
good.
The
American
Economic
Review,
81,
962-970.
Deaton,
A.,
&
Muellbauer,
J.
(1980).
Economics
and
consumer
behavior.
Cambridge:
Cambridge
University
Press.
Elsmore,
T.
F.,
Fletcher,
G.
V.,
Conrad,
D.
G.,
&
Sodetz,
F.
J.
(1980).
Reduction
of
heroin
intake
in
baboons
by
an
economic
constraint.
Pharmacology
Biochemistry
and
Behavior,
13,
729-731.
Fagerstrom,
K.
O.,
&
Schneider,
N.
G.
(1989).
Mea-
suring
nicotine
dependence:
A
review
of
the
Fagerstrom
tolerance
questionnaire.
Journal
of
Behavioral
Medicine,
12,
159-182.
Federal
Trade
Commission.
(1991).
Report
of
"tar"
and
nicotine
content
in
cigarettes.
Washington,
DC:
Author.
Gilley,
0.
W.,
&
Karels,
G.
V.
(1991).
In
search
of
Giffen
behavior.
Economic
Inquiry,
29,
182-189.
Griffiths,
R.
R.,
Bigelow,
G.
E.,
&
Henningfield,
J.
E.
(1980).
Similarities
in
animal
and
human
drug-taking
behavior.
In
N.
K.
Mello
(Ed.),
Advances
in
substance
abuse:
Behavioral
and
biological
research
(Vol.
1,
pp.
1-
90).
Greenwich,
CT:
JAI
Press.
Griffiths,
R.
R.,
Henningfield,
J.
E.,
&
Bigelow,
G.
E.
(1982).
Human
cigarette
smoking:
Manipulation
of
number
of
puffs
per
bout,
interbout
interval
and
nic-
otine
dose.
Journal
of
Pharmacology
and
Experimental
Therapeutics,
220,
256-265.
Hastjarjo,
T.,
&
Silberberg,
A.
(1992).
Effects
of
re-
inforcer
delays
on
choice
as
a
function
of
income
level.
Journal
of
the
Experimental
Analysis
of
Behavior,
57,
119-125.
Hastjarjo,
T.,
Silberberg,
A.,
&
Hursh,
S.
R.
(1990).
Quinine
pellets
as
an
inferior
good
and
a
Giffen
good
in
rats.
Journal
of
the
Experimental
Analysis
of
Behavior,
53,
263-271.
Henningfield,
J.
E.,
&
Goldberg,
S.
R.
(1983).
Nicotine
as
a
reinforcer
in
human
subjects
and
laboratory
an-
imals.
Pharmacology
Biochemistry
and
Behavior,
19,
989-
992.
Henningfield,
J.
E.,
&
Griffiths,
R.
R.
(1981).
Cigarette
smoking
and
subjective
response:
Effects
of
d-amphet-
amine.
Clinical
Pharmacology
and
Therapeutics,
30,
497-
505.
Henningfield,
J.
E.,
Stitzer,
M.
L.,
&
Griffiths,
R. R.
(1980).
Expired
air
carbon
monoxide
accumulation
I
Herrnstein,
R.
J.,
&
Prelec,
R.
(1992).
Giffen
goods
in
rats:
A
reply
to
Battalio
et
al.
Unpublished
manuscript.
500
R.
J.
DeGRANDPRE
et
al.
and
elimination
as
a
function
of
number
of
cigarettes
smoked.
Addictive
Behaviors,
5,
265-272.
Herrnstein,
R.
J.
(1961).
Relative
and
absolute
strength
of
response
as
a
function
of
frequency
of
reinforcement.
Journal
of
the
Experimental
Analysis
of
Behavior,
4,
267-
272.
Herrnstein,
R.
J.
(1970).
On
the
law
of
effect.
Journal
of
the
Experimental
Analysis
of
Behavior,
13,
243-266.
Herrnstein,
R.
J.,
&
Vaughan,
W.,
Jr.
(1980).
Meliora-
tion
and
behavioral
allocation.
In
J.
E.
R.
Staddon
(Ed.),
Limits
to
action:
The
allocation
of
individual
be-
havior
(pp.
143-176).
New
York:
Academic
Press.
Hughes,
J.
R.
(1989).
Environmental
determinants
of
the
reinforcing
effects
of
nicotine.
Journal
of
Substance
Abuse,
1,
319-329.
Hursh,
S.
R.,
&
Bauman,
R.
A.
(1987).
The
behavior
analysis
of
demand.
In
L.
Green
&
J.
H.
Kagel
(Eds.),
Advances
in
behavioral
economics
(Vol.
1,
pp.
117-165).
Norwood,
NJ:
Ablex.
Iglauer,
C.,
&
Woods,
J.
H.
(1974).
Concurrent
per-
formances:
Reinforcement
by
different
doses
of
intra-
venous
cocaine
in
rhesus
monkeys.
Journal
of
the
Ex-
perimental
Analysis
of
Behavior,
22,
179-196.
Johanson,
C.
E.,
&
Schuster,
C.
R.
(1975).
A
choice
procedure
for
drug
reinforcers:
Cocaine
and
methyl-
phenidate
in
the
rhesus
monkey.
Journal
of
Pharma-
cology
and
Experimental
Therapeutics,
193,
676-688.
Katz,
J.
L.
(1990).
Models
of
relative
reinforcing
efficacy
of
drugs
and
their
predictive
utility.
Behavioural
Phar-
macology,
1,
283-301.
Lea,
S.
E.
G.,
Tarpy,
R.
M.,
&
Webley,
P.
(1987).
The
individual
in
the
economy.
Cambridge:
Cambridge
Uni-
versity
Press.
Pearce,
D.
W.
(Ed.).
(1986).
The
MITdictionary
of
mod-
ern
economics
(3rd
ed.).
Cambridge,
MA:
MIT
Press.
Samuelson,
P.
A.,
&
Nordhaus,
W.
D.
(1985).
Econom-
ics
(12th
ed.).
New
York:
McGraw
Hill.
Silberberg,
A.,
Warren-Boulton,
F.
R.,
&
Asano,
T.
(1987).
Inferior-good
and
Giffen-good
effects
in
mon-
key
choice
behavior.
Journal
of
Experimental
Psychology:
Animal
Behavior
Processes,
13,
292-301.
West,
R.,
&
Grunberg,
N.
E.
(1991).
Implications
of
tobacco
use
as
an
addiction.
British
Journal
of
Addiction,
86,
485-488.
Williams,
B.
A.
(1988).
Reinforcement,
choice,
and
re-
sponse
strength.
In
R.
C.
Atkinson,
R.
J.
Herrnstein,
G.
Lindzey,
&
R.
D.
Luce
(Eds.),
Stevens'
handbook
of
experimental
psychology:
Vol.
2.
Learning
and
cognition
(2nd.
ed.,
pp.
167-244).
New
York:
Wiley.
Young,
A.
M.,
&
Herling,
S.
(1986).
Drugs
as
rein-
forcers:
Studies
in
laboratory
animals.
In
S.
R.
Gold-
berg
&
I.
P.
Stolerman
(Eds.),
Behavioral
analysis
of
drug
dependence
(pp.
9-67).
Orlando,
FL:
Academic
Press.
Received
May
4,
1992
Final
acceptance
January
15,
1993
