Vol.
49,
No. 3
May,
1952
Psychological
Bulletin
THE
NATURE
AND
MEASUREMENT
OF
MEANING
1
CHARLES
E.
OSGOOD
University
oj
Illinois*
The
language process
within
an
individual
may be
viewed
as a
more
or
less continuous interaction between
two
parallel systems
of
behavioral
organization:
sequences
of
central
events
("ideas")
and
sequences
of
instrumental
skills,
vocalic, gestural,
or
orthographic, which constitute
the
communicative product.
A
communicator vocalizes,
"It
looks like
rain today;
I'd
better
not
wash
the
car." This output
is a
sequence
of
skilled
movements, complicated
to be
sure,
but not
different
in
kind
from
tying one's shoes. Even
the
smallest units
of the
product,
phonetic
elements
like
the
initial
"l"-sound
of
"looks,"
result
from
precisely pat-
terned muscle movements.
The
organization
of
these movements into
word-units
represents
skill
sequences
of
relatively
high
predictability;
certain longer period sequences involving syntactical order
are
also rel-
atively predictable
for a
given language system.
But
execution
of
such
sequences brings
the
communicator repeatedly
to
what
may be
called
"choice-points"—points
where
the
next
skill
sequence
is
not
highly
predictable
from
the
objective communicative product
itself.
The de-
pendence
of
"I'd better
not
wash
the
car" upon "looks
like
rain
today,"
the
content,
of the
message, reflects determinants within
the
semantic
system
which
effectively
"load"
the
transitional probabilities
at
these
choice-points.
It is the
communicative product,
the
spoken
or
written words which
follow
one
another
in
varying orders,
that
we
typically observe. Since
we
are
unable
to
specify
the
stimuli
which
evoke these communicating
reactions—since
it is
"emitted" rather than
"elicited"
behavior
in
Skinner's
terminology
(97)-—-measurements
in
terms
of
rates
of
occur-
rence
and
transitional probabilities (dependence
of one
event
in the
stream upon others)
are
particularly appropriate (cf.,
Miller,
76).
In-
1
The
research
on
which
in
part this report
is
based
is
being supported
jointly
by the
University
Research Board
of the
University
of
Illinois
and the
Social Science Research
Council.
Grateful acknowledgment
is
made
to
both institutions
for
their assistance.
2
Institute
of
Communications Research
and
Department
of
Psychology.
197
198
CHARLES
E.
OSGOOD
terest
ma}'
be
restricted
to the
lawfulness
of
sequences
in the
observable
communicative product
itself,
without regard
to the
semantic parallel.
This
is
traditionally
the field of the
linguist,
but
even here
it has
proved
necessary
to
make some assumptions
about
meaning (cf.,
Bloomfield,
4). On the
other
hand,
one may be
specifically
interested
in the
semantic
or
ideational level. Since
he is
presently unable
to
observe
this
level
of
behavior directly,
he
must
use
observable characteristics
of the
com-
municative product
as a
basis
for
making inferences about what
is
going
on at the
semantic level.
He may use
sequential orderliness
in the
product
to
draw conclusions
as to
semantic orderliness
in the
speaker's
or
writer's mediation processes
(i.e.,
which
"ideas"
tend
to go
together
in
his
thinking with
greater
than
chance
probabilities).
Or he may
wish
to
study
the
ways
in
which
central,
semantic processes vary
from
con-
cept
to
concept,
from
person
to
person,
and so on. It is the
problem
of
measuring meaning
in
this
latter
sense which
will
be
discussed
in the
present paper.
Before
inquiring into
the
measurement
of the
meaning
of
signs,
for
which there
are no
accepted, standardized techniques available,
we
may
briefly
mention certain
fairly
standard
methods
for
measuring
the
comparative
strength
of
verbal
habits.
Thorndike
and his
associates
(102,
103) have made extensive
frequency-of-usage
counts
of
words
in
English;
that
this
method
gets
at the
comparative habit strengths
of
word
skill sequences
is
shown
by the
fact
that
other measures
of re-
sponse strength, such
as
latency
and
probability within
the
individual
(Thumb
and
Marbe,
106;
Cason
and
Cason,
19),
are
correlated
with
frequency-of-usage.
Zipf
(117, 118,
and
elsewhere)
has
described
in-
numerable instances
of the
lawfulness
of
such habit-strength measures.
Whether samples
be
taken
from
Plautine Latin, newspaper English,
or the
English
of
James Joyce
in his
Ulysses,
a
fundamental regularity
is
found,
such
that
frequency
of
occurrence
of
particular words bears
a
linear
relation
to
their rank order
in
frequency,
when plotted
on
double-
log
paper
(Zipf's
Law). Measurement
of flexibility or
diversity
in
com-
municative products
is
given
by the
type-token ratio
(TTR):
with each
instance
of any
word counting
as a
token
and
each
different
word
as a
type,
the
greater
the
ratio
of
types
to
tokens
the
more varied
is the
con-
tent
of a
message.
This
measure
can be
applied comparatively
to
differ-
ent
forms
of
material,
different
kinds
of
individuals,
and so
forth
(cf.,
Carroll,
16, 17;
Johnson,
45;
Chotlos, 20), provided
the
sizes
of
samples
are
constant.
One may
also count
the
ratios
of
adjectives
to
verbs
(Boder,
5), the
frequencies
of
different
pronouns,
intensives,
and so
forth
(cf.,
Johnson,
45).
NATURE
AND
MEASUREMENT
OF
MEANING
199
Although
the
above measures
get at the
comparative strengths
of
verbal skill sequences
per se
(i.e., without regard
to
meaning), this
is
not a
necessary restriction. Frequency counts
of
this
type
can be
applied
to
semantic habit strengths
as
well. Skinner (96)
has
shown
that
a
similar
lawfulness
applies
to the
frequencies
of
"free" associations
in the
Kent-
Rosanoff
tests.
When
frequencies
of
particular associates
to
given stimu-
lus
words
for a
group
of
subjects
are
plotted against their rank order
in
frequency,
a
straight-line
function
on
double-log paper results
(Zipf's
Law).
In
other words, associations
at the
semantic level appear
to be
organized
in
such
a way
that
few
have very high probability
of
occur-
rence
and
many have
low
probabilities
of
occurrence.
Bousfield
and his
collaborators
(7, 8, 9, 10, 11)
have described
a
sequential association
method
for
getting
at
comparative semantic
habit
strengths. When sub-
jects associate successively
from
the
same
"pool,"
e.
g.,"names
of
four-
legged animals,"
(a) the
rate
of
successive associates shows
a
negatively
accelerated
curve,
(b)
varying
in its
constants with certain characteris-
tics
of
materials
and
subjects,
(c)
the
order
of
appearance
of
particular
associates
in
individuals being predictable
from
the
frequency
of
usage
in
the
group,
and
(d)
distortions
in the
function
being related
to
partic-
ular
transitional probabilities among associates, i.e., clustering.
Useful
though
these measures
are for
many purposes, they
do not get at
mean-
ing.
The
fact
that
"dog"
has a
higher probability
of
occurrence
in se-
quential
association than
"otter"
says
nothing whatsoever about
the
differences
in
meaning
of
these
two
signs.
An
extensive survey
of the
literature
fails
to
uncover
any
generally
accepted, standardized method
for
measuring meaning. Perhaps
it is
because
of the
philosophical haziness
of
this
concept, perhaps because
of
the
general
belief
that
"meanings"
are
infinitely
and
uniquely vari-
able,
or
perhaps because
the
word "meaning"
as a
construct
in our
language
connotes mental
stuff,
more akin
to
"thought"
and
"soul"
than
to
anything
observable—for
some combination
of
reasons
there
has
been little
attempt
to
devise methods here. Nevertheless, whether
looked
at
from
the
viewpoints
of
philosophy
or
linguistics,
from
econom-
ic
or
sociological theory,
or—-interestingly
enough—-from
within
the
core
of
psychological theories
of
individual
behavior,
the
nature
of
meaning
and
change
in
meaning
are
found
to be
central issues.
The
proposals
to be
made
in the
latter
portion
of
this
paper
are
part
of a
program
aimed
at the
development
of
objective methods
of
measuring
meaning.
Beyond obvious social
implications,
it is
felt
that
this direction
of
research
is a
logical extension
of
scientific
inquiry into
an
area
gener-
ally
considered immune
to its
attack.
200
CHARLES
E.
OSGOOD
THEORIES
OF
MEANING
Not all
stimuli
are
signs.
The
shock which galvanizes
a rat
into
vigorous escape movements usually does
not
stand
for
anything other
than
itself,
nor
does
the
pellet
of
food
found
at the end of a
maze,
nor a
hammer
in
one's hand
or a
shoe
on
one's
foot.
The
problem
for any
meaning
theorist
is to
differentiate
the
conditions under which
a
pattern
of
stimulation
is a
sign
of
something else
from
those
conditions where
it
is
not.
This
certainly seems simple enough,
yet it has
troubled philos-
ophers
for
centuries.
By
stating
the
problem somewhat
formally,
the
chief
differences
between several conceptions
of the
sign-process
can
be
made evident:
let
S
=
object
=
any
pattern
of
stimulation
which
evokes reactions
on the
part
of an
organism,
and
D0
=
sign
=any
pattern
of
stimulation
which
is not
this
5
but yet
evokes
reactions
relevant
to
5—conditions
under which
this
holds being
the
problem
for
theory.
The
definition
of
S
is
broad
enough
to
include
any
pattern
of
stimula-
tion which elicits
any
reaction
from
an
organism. Although
one
usually
thinks
of
"objects"
as
those things denoted
by
signs, actually
any
pattern
of
stimulation—a
gust
of
northerly wind against
the
face,
the
sensations
we
call "belly-ache,"
the
sensations
of
being rained
upon—
is
an
"object"
at
this level
of
discourse.
One
sign
may be the
"object"
represented
by
another sign,
as
when
the
picture
of an
apple
is
called
"DAX"
in
certain experiments.
The
definition
of
CD
is
purposely
left
incomplete
at
this
point, since
it
depends upon one's conception
of the
nature
of the
sign-process.
We
may
start
a
logical analysis
of the
problem with
a
self-evident
fact:
the
pattern
of
stimulation which
is the
sign
is
never
identical with
the
pattern
of
stimulation which
is the
object.
The
word
"hammer"
is not the
same stimulus
as is the
object hammer.
The
fofmer
is a
pattern
of
sound waves having characteristic oscillations
in
frequency
and
inten-
sity;
the
latter,
depending upon
its
mode
of
contact,
may be a
visual
form
having characteristic color
and
shape,
a
pattern
of
tactual
and
proprioceptive sensations,
and so on.
Similarly,
the
buzzer
in a
typical
rat
experiment
is not
identical
as a
form
of
stimulation with
the
shock
which
it
comes
to
signify.
Yet
these
signs—the
word
"hammer"
and the
buzzer—do
elicit behaviors which
are in
some manner
relevant
to the
objects they
signify,
a
characteristic
not
shared with
an
infinite
number
of
other
stimulus
patterns
that
are not
signs
of
these
objects.
In
simplest
terms, therefore,
the
question
is:
under what conditions
does
something
NATURE
AND
MEASUREMENT
OF
MEANING
101
which
is
not,
an
object
become
a
sign
of
that
object?
According
to the way
in
which this question
is
answered
we may
distinguish several theories
of
meaning.
Mentalistic
View
The
classic interpretation derives directly
from
the
natural
philos-
ophy
of
Western
culture,
in
which
the
dualistic connotations
of
lan-
guage
dictate
a
correlation between
two
classes
of
events,
material
and
nonmaterial.
Since meanings
are
obviously
"mental"
events
and the
stimuli
representing objects
and
signs
are
obviously "physical"
events,
any
satisfying theory
of
meaning must
specify
interrelation between
these levels
of
discourse.
At the
core
of all
mentalistic views, therefore,
we
find an
"idea"
as the
essence
of
meaning;
it is
this mental event
which
links
or
relates
the two
different
physical events, sign
and
object.
The
word
"hammer"
gives rise
to the
idea
of
that
object
in the
mind;
conversely, perception
of the
object hammer gives rise
to the
same idea,
which
can
then
be
"expressed"
in
appropriate signs.
In
other words,
something which
is not the
object
becomes
a
sign
of
that
object
when
it
gives
rise
to the
idea associated with that
object.
Probably
the
most sophisti-
cated expression
of
this view
is
given
by
Ogden
and
Richards (82)
in
their book,
The
Meaning
of
Meaning.
Most readers
will
recall their
triangular diagram
of the
sign-process:
the
relation between symbol
and
referent
(the
base
of
their triangle)
is not
direct
but
inferred,
medi-
ated
through mental
"thought"
or
"interpretation" (the third corner
of
their triangle),
Substitution
View
Naive application
of
Pavlovian
conditioning principles
by
early
behaviorists like Watson
led to the
theory that signs achieve their
meanings
simply
by
being conditioned
to the
same reactions originally
made
to
objects.
This,
in
essence,
is the
view
one
encounters
in
many
introductory
texts
in
general psychology.
An
object evokes certain
be-
havior
in an
organism;
if
another
pattern
of
stimulation
is
consistently
paired with
the
original
object,
it
becomes conditioned
to the
same
re-
sponses
and
thus gets
its
meaning.
The
object
is the
unconditioned
stimulus
and the
sign
is the
conditioned
stimulus,
the
latter
merely
being
substituted
for the
former.
The
definition
of the
sign-process here
is
that
whenever
something which
is not the
object
evokes
in an
organism
the
same reactions
evoked
by the
object,
it is a
sign
of
that
object.
The
very
simplicity
of
this
theory highlights
its
inadequacy.
Signs
almost never
evoke
the
same overt responses
as do the
objects they represent.
The
word
FIRE
has
meaning
to the
reader without sending
him
into head-
long
flight.
Nevertheless,
this represents
a first
step
toward
a
behavioral
interpretation
of the
sign-process.
