ArticlePDF Available

Pigeons’ discrimination of painting by Monet and Picasso

Authors:

Abstract and Figures

Pigeons successfully learned to discriminate color slides of paintings by Monet and Picasso. Following this training, they discriminated novel paintings by Monet and Picasso that had never been presented during the discrimination training. Furthermore, they showed generalization from Monet's to Cezanne's and Renoir's paintings or from Picasso's to Braque's and Matisse's paintings. These results suggest that pigeons' behavior can be controlled by complex visual stimuli in ways that suggest categorization. Upside-down images of Monet's paintings disrupted the discrimination, whereas inverted images of Picasso's did not. This result may indicate that the pigeons' behavior was controlled by objects depicted in impressionists' paintings but was not controlled by objects in cubists' paintings.
Content may be subject to copyright.
JOURNAL
OF
THE
EXPERIMENTAL
ANALYSIS
OF
BEHAVIOR
PIGEONS'
DISCRIMINATION
OF
PAINTINGS
BY
MONET
AND
PICASSO
SHIGERU
WATANABE,
JUNKO
SAKAMOTO,
AND
MASUMI
WAKITA
KEIO
UNIVERSITY
Pigeons
successfully
learned
to
discriminate
color
slides
of
paintings
by
Monet
and
Picasso.
Following
this
training,
they
discriminated
novel
paintings
by
Monet
and
Picasso
that
had
never
been
presented
during
the
discrimination
training.
Furthermore,
they
showed
generalization
from
Monet's
to
Ce-
zanne's
and
Renoir's
paintings
or
from
Picasso's
to
Braque's
and
Matisse's
paintings.
These
results
suggest
that
pigeons'
behavior
can
be
controlled
by
complex
visual
stimuli
in
ways
that
suggest
categorization.
Upside-down
images
of
Monet's
paintings
disrupted
the
discrimination,
whereas
in-
verted
images
of
Picasso's
did
not.
This
result
may
indicate
that
the
pigeons'
behavior
was
controlled
by
objects
depicted
in
impressionists'
paintings
but
was
not
controlled
by
objects
in
cubists'
paintings.
Key
words:
stimulus
control,
concept,
pattern
discrimination,
vision,
key
peck,
pigeon
When
we
see
paintings
by
Picasso
and
Mo-
net,
we
can
with
some
accuracy
recognize
which
is
Picasso's
and
which
is
Monet's,
even
if
we
have
never
seen
the
particular
paintings
be-
fore.
There
are
many
possible
cues
for
this
discrimination,
such
as
color,
style
of
brushing,
favorite
subjects,
and
so
on,
but
no
single
fea-
ture
differentiates
each
artist.
It
is
also
clear
that
we
have
acquired
such
visual
concepts
of
paintings
of
Picasso
and
Monet
by
experience.
Can
pigeons
discriminate
paintings
of
one
art-
ist
from
those
of
another
artist?
If
they
can,
do
they
also
show
generalization
to
paintings
of
other
artists
belonging
to
the
same
group,
such
as
an
impressionist
or
a
cubist?
Porter
and
Neuringer
(1984)
reported
successful
learning
of
musical
discrimination
of
Bach
and
Stravinsky
by
pigeons.
Can
pigeons
discrimi-
nate
visual
arts
also?
Birds
have
excellent
visual
ability
compa-
rable
to
that
of
humans,
and
there
have
been
many
experimental
studies
showing
acquisi-
tion
of
visual
concepts
in
birds.
Since
Herrn-
stein
and
Loveland
(1964)
successfully
trained
pigeons
to
respond
to
color
slides
on
which
a
human
being
appeared
and
not
to
respond
to
those
without
a
human,
there
have been
many
studies
demonstrating
learning
to
discriminate
natural
concepts
(e.g.,
Cerella,
1979;
Herrn-
The
authors
wish
to
express
their
gratitude
to
R.
J.
Herrnstein
for
encouragement
and
comments
on
this
re-
search.
This
research
was
supported
by
Grant-in-Aid
for
Promotion
of
Sciences
05206113.
Address
correspondence
and
reprint
requests
to
Shigeru
Watanabe,
Psychology
Department,
Keio
University,
Mita
2-15-45,
Minato-ku,
Tokyo,
Japan.
stein
&
de
Villiers,
1980;
Herrnstein,
Love-
land,
&
Cable,
1976;
Roberts
&
Mazmanian,
1988;
Watanabe,
Yamasita,
&
Wakita,
1993),
artificial
concepts
(Bhatt,
Wasserman,
Reyn-
olds,
&
Knauss,
1988;
Watanabe,
1991),
and
symmetry
of
objects
(Delius
&
Habers,
1978).
Most
of
these
natural-concept
experiments
used
a
slide
projector
as
the
stimulus-presen-
tation
device,
and
pigeons
showed
transfer
of
discrimination
of
photographs
to
real
objects
and
of
real
objects
to
photographs
(Watanabe,
1993).
Representational
paintings
have
fea-
tures
similar
to
photographs,
but
paintings
patterned
after
impressionism
are
not
precise
reflections
of
the
real
world.
They
often
are
considered
to
be
a
reflection
of
the
artist's
sub-
jective
world.
We
can,
however,
identify
"ob-
jects"
in
the
paintings
by
Monet,
Renoir,
and
Cezanne.
In
other
words,
we
find
a
relation
between
these
paintings
and
real
objects.
How-
ever,
such
a
relation
is
often
weak
in
the
paint-
ings
by
Picasso,
Matisse,
and
Braque.
Realism
is
relevant
only
for
a
perceiver
who
can
see
a
painting
as
a
representation
of
a
three-dimen-
sional
world.
If
realism
makes
a
difference
to
a
pigeon,
we
can
presume
that
it
can
see
a
painting
as
a
representation
of
a
three-dimen-
sional
world.
EXPERIMENT
1
In
this
experiment
pigeons
were
trained
on
a
discrimination
between
photographs
or
vi-
deos
of
paintings
by
Monet
and
those
of
Pi-
casso.
The
paintings
differed
in
their
color,
sharpness
of
contour,
and
objects.
Potential
cues
for
discrimination
were
examined
by
as-
165
1995,
639
165-174
NUMBER
2
(MARCH)
SHIGERU
WATANABE
et
al.
sessing
effects
of
distortion,
such
as
left-right
reversal
and
upside-down
reversal.
Keller
and
Schoenfeld
(1950)
defined
a
concept
as
a
gen-
eralization
within
a
class
of
stimuli
and
a
dis-
crimination
between
the
classes.
The
gener-
alization
of
the
discrimination
of
paintings
by
Monet
and
Picasso
to
novel
paintings
of
these
artists
and
those
of
other
artists
was
tested.
METHOD
Subjects
Eight
experimentally
naive
pigeons
(Co-
lumba
livia)
were
used
for
the
present
exper-
iment.
They
were
individually
housed
in
stain-
less
steel
cages
and
were
maintained
at
about
80%
of
their
free-feeding
body
weights.
Apparatus
The
experimental
chambers
were
two
iden-
tical
operant
conditioning
chambers
(30
cm
by
30
cm
by
30
cm)
with
a
rectangular
glass
peck-
ing
key
(5
cm
by
7
cm).
The
key
could
be
activated
by
a
force
of
0.2
N.
A
frosted
glass
screen
(5
cm
by
7
cm)
was
attached
1
cm
behind
the
key.
The
key
was
mounted
20
cm
above
the
floor
and
13
cm
above
the
aperture
of
a
food
hopper.
Stimuli
were
projected
on
the
screen
by
a
slide
projector
(Super
Cabin
2)
in
one
chamber
and
by
a
video
projector
(Phillips
LCP5000)
in
another
chamber.
A
microcomputer
(Sanyo
MSX)
controlled
the
experiment.
The
projected
stimuli
provided
the
only
illumination
in
the
chamber.
The
cham-
ber
was
not
sound
isolated,
but
white
noise
(70
dB)
was
continuously
broadcast.
Stimuli
Two
different
sets
of
stimuli,
taken
from
picture
books,
were
used
as
training
stimuli
(Table
1).
Each
set
consisted
of
10
different
paintings
by
Monet
and
10
by
Picasso.
Set
A
was
used
for
training
with
the
slide
projector,
and
Set
B
was
used
for
training
with
the
video
projector.
"Typical"
paintings
were
selected
as
stimuli.
"Atypical"
paintings
(such
as
those
of
Picasso's
blue
period)
were
not
used
as
stim-
uli.
Table
2
lists
stimuli
used
for
the
gener-
alization
test.
Three
novel
paintings
by
Monet
and
Picasso
and
three
each
by
Cezanne,
Braque,
and
Delacroix
were
used
as
testing
stimuli
after
training
with
Set
A.
Four
novel
Monet
paintings,
four
new
Picasso
paintings,
and
four
each
by
Renoir,
Matisse,
and
De-
lacroix
were
presented
in
the
generalization
test
after
training
with
Set
B.
Procedure
The
pigeons
were
first
trained
to
peck
the
key
illuminated
by
a
projector
lamp
without
any
painting
stimulus.
Then
they
were
divided
into
two
groups
of
4
birds
each.
In
the
Monet
S+
group,
responses
to
paintings
of
Monet
were
reinforced
by
4-s
access
to
a
feeder
con-
taining
hemp
seeds
and
responses
to
paintings
of
Picasso
were
extinguished,
whereas
in
the
Picasso
S+
group
paintings
of
Picasso
were
associated
with
reinforcement
and
those
of
Monet
were
not.
Two
different
sets
of
stimuli
were
used.
Two
birds
in
each
group
received
training
with
Set
A,
and
the
other
2
were
trained
with
Set
B.
A
training
session
consisted
of
20
randomly
ordered
presentations
of
each
painting
once,
each
lasting
30
s,
separated
by
a
5-s
blackout
period.
During
presentation
of
S+,
reinforce-
ment
was
available
on
a
variable-interval
(VI)
30-s
schedule,
whereas
no
reinforcement
was
available
during
S-
periods
(mult
VI
30-s
ext).
The
subjects
received
one
training
session
every
day.
This
discrimination
training
continued
until
the
subjects
showed
a
discrimination
ratio
above
90%,
calculated
by
dividing
the
number
of
re-
sponses
to
S
+
by
the
total
number
of
responses,
summed
over
two
successive
sessions.
Then
the
following
tests
were
carried
out
under
extinc-
tion
(i.e.,
pecking
did
not
activate
the
food
hopper
during
the
tests).
The
subjects
received
at
least
two
sessions
of
discriminative
training
between
tests.
If
the
subjects
did
not
show
at
least
90%
correct
on
the
two
sessions,
addi-
tional
training
was
given
until
the
subjects
reached
the
criterion
again.
Test
1.
Color
of
paintings
by
Monet
might
differ
from
those
by
Picasso.
Differences
in
color
thus
might
serve as
cues
for
discrimi-
nation.
In
this
test,
monochromatic
pictures
of
all
training
paintings
were
used
to
examine
the
possibility
that
color
cues
controlled
the
discrimination.
The
order
and
period
of
stim-
ulus
presentation
were
similar
to
those
in
daily
training
sessions.
Monochromatic
slides
were
used
for
Set
A,
and
monochromatic
images
were
produced
by
tuning
the
video
projector
for
Set
B.
Test
2.
Most
of
the
paintings
by
Picasso
have
sharp
contours,
but
most
of
those
by
Monet
166
DISCRIMINATION
OF
PAINTINGS
11
o0
-1-00
-
*"0
"
n
u-
n
t
cnL
-
-
O
el
cl c4
cn
b
*
"
N
M
t-
M
O
bF
..000
N
-
O%O
",
*
G
.,
N
1
C.)
^
Y
Xo
so
%O
tl
t-
tl
(O
co
N-
oD
a
o
o-4--WN
eq
N
_
oo
co
co
co
oo
ao
oo
co
oi
a,
os
Co
ofi
ok
ON
o.
ok
0,
o'
3CU
1
0
E
os
°~
.D
O e X
@
:N
r-4
X
Xo
o
X
b
o
O_
_
N
t
+
*n
CO
*_
'O
Q
co
4
B
&>
_
o
on
n
.
r
¢Y0
U~~~~~~~~c
B~~~~~~
~
~~~~
4i
co
co
,,,
y
oYE.6Y
;W
CIO_
C.)
C-h
m
~~~~T
Q
zCU
i
bo
~ ~
0
n0
rC
O,
n
0C4VN
00c
v-
v
-
N
c
q
N
W
-
CUCUN
en
eU
Cl
ncnCOU
U)
cn0
)c
n0
nCf
/
nC)c
~
~
~
~ ~ ~ ~ ~ ~ ~ ~ ~ ~
C
NO)
CU~
~
~~~~~'
0~~~~~~~~~0
C(
167
ll
SHIGERU
WATANABE
et
al.
Table
2
Paintings
used
for
tests.
S
and
R
indicate
catalogue
numbers
as
Table
1.
M
indicates
catalogue
number
in
Matisse
by
J.
Guillaud
and
M.
Guillaud
(Guillaud
Edition,
1987,
Paris).
Test
stimuli
after
training
with
Set
A
Test
stimuli
after
training
with
Set
B
New
Monet
La
Grenouillere
Lady
with
a
parasole
Water
lily
New
Picasso
Dance
Woman
looking
at
the
glass
Still
life
with
an ox
head
Cezanne
Sitting
man
Still
life
with
onions
Big
water
bathing
Braque
Female
musician
Still
life
with
"le
Jour"
An
easel
and
a
woman
Delacroix
Still
life
with
a
lobster
July
28th
Atelier
New
Monet
1869
S8
Le
Grenouillere
1886
S23
The
Thames
ant
the
capital
1914
S31
Church
Palazzo
da
mula
in
Venezia
IoQ-,
Iq1
New
Picasso
1937
S25
1942
S27
1898
S17
1895
S23
1898
S27
1917
S13
1929
S24
1936
S27
1827
S5
1830
S8
1830
S9
Donna
con
ventaglio
Natura
morta
spagnola
Foglio
di
musica
e
chitarra
Arlecchino
Matisse
Notre-dame
La
Tovaglia
Blouse
roumaine
Apricot
Renoir
La
senna
ad
argenteuil
Canottieri
ad
argenteuil
Donna
alla
grenouilere
Frutta
del
meridione
Delacroix
Still
life
with
lobster
Atelier
July
28th
Shopin
1869
S8
1871
S1l
1883
S22
1908
S29
1909
R224
1912
R471
1912
R577
1913
R615
1902
S3
1908
R107
1940
M413
1948
S30
1883
R96
1883
R97
1879
R373
1881
R486
1827
S5
1830
S9
1830
S8
1838
S15
do
not.
In
this
test
all
training
paintings
were
presented
out
of
focus
to
examine
the
role
of
contour.
Other
procedures
of
the
stimulus
pre-
sentation
were
identical
to
those
in
daily
train-
ing
sessions.
In
these
pictures,
two
lines
(0.43
mm
wide)
separated
by
0.43
mm
fused
com-
pletely
on
the
screen.
Test
3.
Training
stimuli
were
used
here
but
three
S
+
and
three
S
-
pictures
were
left-right
reversed
and
another
three
S+
and
three
S-
pictures
were
upside
down.
Six
normal
stimuli
were
also
presented.
The
same
testing
proce-
dure
was
used
here
(i.e.,
each
test
stimulus
was
presented
once
for
30
s).
Test
4.
As
shown
in
Table
2,
two
different
sets
of
stimuli
appeared
in
generalization
tests.
One
set,
which
was
used
after
training
with
Set
A,
consisted
of
three
novel
paintings
by
Monet,
three
novel
paintings
by
Picasso,
and
three
each
by
Cezanne,
Braque,
and
Dela-
croix.
These
15
new
stimuli
and
six
old
stimuli
used
for
discriminative
training
(three
each
by
Picasso
and
Monet)
were
randomly
presented
three
times
each.
The
other
set,
which
con-
sisted
of
four
novel
paintings
each
by
Monet
and
Picasso
and
four
each
by
Renoir,
Matisse,
and
Delacroix,
was
used
after
training
with
Set
B.
These
20
new
stimuli
and
eight
old
stimuli
(four
each
by
Picasso
and
Monet)
ran-
domly
appeared
three
times
each
during
the
tests.
RESULTS
AND
DISCUSSION
All
subjects'
behavior
came
under
stimulus
control.
The
number
of
sessions
to
reach
the
criterion
ranged
from
6
to
22
for
the
Monet
S+
group
and
8
to
24
for
the
Picasso
S+
group.
There
was
no
statistically
significant
difference
in
speed
of
acquisition
between
the
Monet
S
+
and
Picasso
S
+
groups
(two-tailed
t
test,
=
0.86,
df
=
6)
or
between
Set
A
and
Set
B
(t
=
1.64,
df=
6).
During
the
training
period,
the
subjects
re-
sponded
more
often
to
some
paintings
than
to
other
paintings.
There
might
be,
thus,
paint-
ings
that
are
easy
to
discriminate
and
those
168
DISCRIMINATION
OF
PAINTINGS
that
are
hard
to
discriminate.
To
examine
the
role
of
each
painting
in
acquisition
of
discrim-
ination,
total
number
of
responses
emitted
to
each
S+
painting
until
each
subject
reached
the
criterion
were
analyzed.
Because
the
num-
ber
of
sessions
to
the
criterion
differed
among
subjects,
the
rank
order
of
the
responses
to
each
S+
painting
was
used
for
the
analysis.
The
Spearman's
rank-order
correlation
coefficient
(rho)
between
2
subjects
in
the
Monet
S+
group
trained
with
Set
A
was
-.02,
and
that
with
Set
B
was
.06.
The
rho
between
the
2
subjects
in
the
Picasso
S+
group
trained
with
Set
A
was
.19,
and
that
with
Set
B
was
.37.
None
of
the
correlations
was
statistically
sig-
nificant.
There
was
thus
no
systematic
bias
of
responding
caused
by
particular
stimuli,
even
though
individual
subjects
subjects
showed
dif-
ferences
in
responding
to
each
stimulus.
Figure
1
presents
results
of
Tests
1
and
2.
Because
there
were
individual
differences
in
absolute
number
of
pecks,
percentage
of
cor-
rect
responses
is
shown
in
the
figure.
All
birds
maintained
their
discrimination
in
the
mon-
ochromatic-stimulus
test,
although
some
birds
showed
less
than
90%
correct.
Thus,
color
was
not
a
crucial
cue
for
the
painting
discrimina-
tion.
There
was
no
statistically
significant
dif-
ference
between
the
Monet
S+
group
and
the
Picasso
S+
group
(unpaired
two-tailed
t
=
0.176,
df
=
6).
Most
of
the
paintings
of
impressionists
lack
sharp
outlines,
whereas
most
of
the
paintings
of
cubists
have
sharp
contours.
Sharp
contours
that
might
be
a
cue
for
discrimination
were
investigated
in
Test
2.
Although
D24
and
B34
in
the
Picasso
S
+
group
showed
less
than
90%
correct
responding
in
the
test
of
focus,
other
birds
showed
more
than
90%
correct
respond-
ing.
D24
saw
Set
A,
and
B34
saw
Set
B.
Birds
can
show
individual
differences
in
selective
stimulus
control in
discriminations
of
com-
pound
stimuli
consisting
of
shape
and
color
cues
(Reynolds,
1961)
and
in
discrimination
of
more
complicated
stimuli
(Watanabe
et
al.,
1993).
There
was
no
statistically
significant
difference
in
percentage
correct
between
the
Monet
S+
and
Picasso
S+
groups
in
the
test
of
focus
(unpaired
two-tailed
t
=
0.635,
df
=
6).
Results
of
Tests
1
and
2,
together
with
the
analysis
of
the
categories
of
subjects
of
paint-
ings,
suggest
that
neither
category,
color,
nor
edge
sharpness
uniquely
controlled
the
dis-
crimination.
b
LU
0
z
LU.
LUJ
0l.
N
m
80
--
m
Lf)
91
60
n0
40-
20-
-111
11
11