The Psychological Foundations of the Affective Lexicon

Abstract

Subjects rated their confidence that each word from a set of 585 words referred to an emotion. As a strategy for discriminating words that refer to genuine emotions from words that refer to other kinds of conditions, ratings were collected in two different linguistic contexts: first, in the context of
feeling something and second, in the context of
being something. We hypothesized that words that referred to genuine emotions would be judged as such when presented in the context of feeling or being (e.g.,
feeling angry and
being angry should both be rated as emotions). Words not referring to genuine emotions, however, were expected to show one of several other patterns. For example, words such as
abandoned, which refer to objective states of the world, were expected to be rated as emotions in the feeling context but not in the being context. A discriminant analysis showed that such patterns could be used to distinguish the categories of a taxonomy of psychological conditions that Ortony, Clore, and Foss (1987) have proposed. The most discriminable categories were the four classes of affective, cognitive, external, and bodily conditions. (PsycINFO Database Record (c) 2012 APA, all rights reserved)

Full text

Journal
of
Personality
and
Social
Psychology
1987,
Vol.
53,
No.
4.751-766
Copyright
1987
by the
American
Psychological
Association,
Inc.
0022.3514/87/J00.75
The
Psychological
Foundations
of the
Affective
Lexicon
Gerald
L.
Clore,
Andrew
Ortony,
and
Mark
A.
Foss
University
of
Illinois
at
Urbana-Champaign
Subjects
rated
their confidence that each word
from
a set
of
585
words referred
to an
emotion.
As a
strategy
for
discriminating words that
refer
to
genuine
emotions
from
words
that
refer
to
other kinds
of
conditions,
ratings were
collected
in two
different
linguistic
contexts:
first, in the
context
of
feeling
something
and
second,
in the
context
of
being
something.
We
hypothesized
that
words that referred
to
genuine emotions would
be
judged
as
such
when
presented
in the
context
of
feeling
or
being (e.g.,
feeling
angry
and
being
angry
should both
be rated as
emotions). Words
not
referring
to
genuine
emotions,
however,
were
expected
to
show
one of
several other patterns.
For
example, words such
as
abandoned,
which
refer
to
objective
states
of
the
world, were expected
to be
rated
as
emotions
in the
feeling
context
but not in the
being context.
A
discriminant analysis showed that such patterns could
be
used
to
distinguish
the
categories
of a
taxonomy
of
psychological conditions
that
Ortony, Clore,
and
Foss
(1987)
have
proposed.
The
most
discriminable
categories
were
the
four
classes
of
affective,
cognitive,
external,
and
bodily
conditions.
When
is a
feeling
an
emotion?
One can
experience
many
kinds
of
feelings;
one can
feel
sleepy,
certain,
proud,
or
aban-
doned.
One can
even
feel
like eating Chinese
food.
All of
these
are
legitimate uses
of the
terra
feeling,
but not all of
them
refer
to
emotions.
Before
an
adequate theory
of
emotion
can be
de-
veloped,
some criteria must
be
established
for
separating
emo-
tional
from
nonemotional
feelings.
Without such
criteria,
one
has no way of
knowing
to
what
the
theories
refer.
Psychologists
have
generally
not
concerned themselves
with
this
problem,
of-
ten
using
as
stimuli
in
their studies words whose status
as
emo-
tion
terms
is
questionable.
For
example,
it
does
not
seem unrea-
sonable
to
question
the
status
of the
following
terms:
sleepy,
tired,
and
relaxed
(Russell,
\9&0);
puzzled,
curious,
and
recep-
tive
(Plutchik,
1962);
boredom,
impatience,
and
inspiration
(Davitz,
1969);
and
luckiness,
conflict,
and
rectitude
(Abelson,
1983).'
Do
these words
refer
to
emotions?
To
address this
ques-
tion,
one
needs
a
principled criterion
for
distinguishing
emo-
tions
from
nonemotions,
because
in the
absence
of
such
a
crite-
rion, theoretical
and
empirical treatments
of
emotions
are
likely
to be
invalid
and
misleading.
Some evidence
of
this
is
reported
in
Morgan
and
Heise
(1986),
who
found
that studies
using
only
words that seem
to be
good examples
of
emotions
yield
rather
different
results
from
studies using
a
less strict crite-
rion
for
selecting emotion terms
from
the
general class
of
affec-
tive
words.
Our
study
was
designed
to
test
the
validity
of a
tax-
This
work
was
supported
in
part
by a
grant
from
the
National Science
Foundation,
BNS
83-18077,
awarded
to
Gerald
L.
Clore
and
Andrew
Ortony
and in
part
by a
contract
from
the
National
Institute
of
Educa-
tion under Contract
400-81
-0030,
awarded
to the
Center
for
the
Study
of
Reading
at the
University
of
Illinois.
We
wish
to
thank Terence
J.
Turner
for his
help
on
various
aspects
of
this work.
Correspondence concerning this article should
be
addressed
to
Ger-
ald L.
Gore.
Department
of
Psychology, University
of
Illinois,
603
East
Daniel.
Champaign,
Illinois
61820.
onomy
of
affective
concepts
and to
explore
the
effectiveness
of
a
criterion that distinguishes emotions
from
other concepts
in
the
affective
lexicon.
Before
going
any
further,
we
need
to
clarity
a
terminological
issue
that
might
otherwise
be a
source
of
misunderstanding.
Throughout this article
we use the
term
affective
to
refer
to the
positive
or
negative evaluation,
or
valence, inherent
in the
meaning
of a
term. More
generally,
we use the
phrase
affective
lexicon
to
refer
to
that subset
of
words
in a
language
that
are
about
affect
or
affective
conditions.
Many,
but by no
means all,
of
the
words
in the
affective
lexicon
refer
to
emotions.
Affect,
therefore,
is
being
treated
as a
more general concept than
is
emotion:
All
emotions
are
affective,
but not all
affective
condi-
tions
are
emotions.
Other investigators
have
studied
the
structure
of the
affective
lexicon, although usually
with
different
goals than ours (e.g.,
Averill,
1975; Bush, 1973;
Dahl
&
Stengel,
1978;
Davitz,
1969;
Russell, 1980).
Typically,
they
have
applied multidimensional
scaling
and
factor
analytic procedures
to
presumed emotion
words
in
attempts
to
discover
the
structure
of
emotions. That
is,
they
have
assumed that their
stimulus
words
refer
to
emo-
tions
and
have
used such procedures
to
determine dimensions
in
terms
of
which
emotions
can be
discriminated
from
one an-
other.
But
what
would
be the
consequences
if
some
of the
stimu-
lus
words used
in
such studies
were
in
fact
not
emotion words
at
all?
An
obvious consequence
is
that this
would
reduce
one's
confidence
that
the
resulting dimensions
or
factors
described
the
structure
of
emotions properly.
We
see
other problems
with
the
kind
of findings
that result
from
many
scaling studies
of
emotions. Because
of the
nature
of
the
judgment task, such studies
often
yield
rather general
di-
mensions
that
are not
particularly
informative
with
respect
to
1
We
should note that Abelson
(1983)
acknowledged that
rectitude
may
not
be
a
good example
of
an
emotion.
751
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

752
G.
CLORE,
A.
ORTONY,
AND M.
FOSS
the
kinds
of
questions that
ought
to
concern emotion theorists.
For
example,
the two
dimensions
(often
interpreted
as
valence
and
arousal)
most
frequently
discovered
by
scaling
procedures
appear
to
have
no
particular relation
to
emotions.
One is
simply
Osgood's
evaluation
dimension
(E),
and the
other
is
easily inter-
preted
as a
combination
of his
potency
(P) and
activity
(A) di-
mensions
(Osgood,
Suci,
&
Tannenbaum,
1957).
But
these,
of
course,
are
universal
dimensions
in
terms
of
which
any
concept
can be
scaled.
If
this
is
correct, then nothing
unique
about
the
nature
of
emotions
has
been discovered unless,
perhaps,
it can
be
shown
that
all and
only
emotions occupy
a
unique
subregion
in
the
space
the
dimensions
define.
However,
Osgood,
May,
&
Miron's
(1975)
data suggested that this
is not the
case.
It is
easy
to find
ratings
of
emotions
in
their data that
are
more similar
to
ratings
of
nonemotions
than
they
are to
ratings
of
other
emo-
tions.
For
example,
the E, P, and A
ratings
for
sympathy
(1.6,
0.6, 0.3)
are
closer
to
those
(or
food
(1.8, 0.5, 0.3) than they
are
for, say,
love
(2.0, 1.2, 0.8);
see
also Carroll
(1959).
Thus,
dimensions
of
this
kind
are so
general that
they
are
quite
unin-
formative
with
respect
to
identifying
features
that distinguish
emotions
from
other things.
They
reveal
no
principled
defini-
tive
differences
between emotions (e.g.,
sympathy)
and
things
having
nothing
whatsoever
to do
with
emotions
(e.g.,
food).
Nor
are
they
informative
with
respect
to
distinguishing
one
type
of
emotion,
say,
anger
(construed
as
representing
all
anger tokens,
such
as
annoyed,
irritated,
furious,
and
enraged)
from
another
type,
say
fear.
Because intensity
is not
likely
to be an
important
feature
for
distinguishing
between
different
emotion types
(as
opposed
to
tokens),
only
valence
would
remain
as a
significant
dimension
in a
typical
two-dimensional
solution (assuming that
an
arousal, that
is,
activity-potency, dimension
reflects
little
more
than
intensity).
It
hardly
seems plausible
to
suppose that
the
psychological
difference
between emotion types
can be
cap-
tured
properly simply
in
terms
of one
dimension (e.g., valence).
For
these
and
similar
reasons,
we
believe
it is
important
to es-
tablish
some criteria
for
distinguishing
emotion terms
from
other terms
in the
affective
lexicon
before
one can
even
hope
to
discover
anything
informative
about emotions
by
using
scaling
methods.
A
Taxonomy
of
Affective
Conditions
In
Ortony,
Clore,
and
Foss
(1987),
we
proposed
an
account
of
the
referential
structure
of
approximately
500
words
in the
affective
lexicon,
using
constructs that
were
independent
of
any
particular theory
of
emotion.
In the
study presented
in
this
arti-
cle,
we
investigated whether subject ratings
of
essentially
the
same
sample
of
words
could
be
used
to
discriminate
the
catego-
ries
of
that taxonomy (see Figure
1).
Figure
1
shows
the
salient
features
(enclosed
in
ellipses)
in
terms
of
which
the
different
psychological conditions
are
distin-
guished.
We
present
only
a
synopsis
of the
distinctions here;
a
more
detailed account
is
presented
in
Ortony
et
al.
(1987).
The
first
main
distinction
is
between words that
refer
to
Internal
Conditions
and
those that
refer
to
External
Conditions.2
Terms
that
refer
to
Internal Conditions
refer
to
conditions
of the
per-
son of
whom
the
term
is
predicated. These
are
typically,
but
not
exclusively,
experienced
states.
Words
that
refer
to
External
Conditions
do not
refer
directly
to
experiences
of the
person
of
whom
they
are
predicated.
Two
kinds
of
External Conditions
can be
distinguished,
namely
Subjective
Evaluations
such
as
sexy,
peculiar,
and
weird,
and
Objective
Descriptions
such
as
alone,
abandoned,
and
welcome.
The
reason
we
call these Exter-
nal
Conditions
is
that
when
one
describes
a
person
as
sexy,
or
as
abandoned,
one is not
referring
directly
to any
particular
experience
of
that
person.3
As
shown
in
Figure
1,
two
types
of
Internal Conditions
are
proposed:
one
comprising
Nonmental
Conditions
and the
other
comprising
Mental
Conditions.
The
Nonmental Conditions
in-
clude
only
Physical
and
Bodily
States,
referred
to by
terms
such
as
aroused,
sleepy,
and
well.
The
Mental Conditions
give
rise
to five
categories,
depending
on
which
of
three major meaning
components,
affect,
behavior,
and
cognition,
are
referentially
focal.
A
referentially
focal
meaning,
component
is one
that con-
stitutes
a
predominant part
of the
reference
of a
term rather
than
something that
is
merely implicated
in its
meaning.
Of the
categories
for
which
affect
is
focal,
we first
identify
a
category
of
Affective
States
in
which
only
affect
is
focal. This category
includes
states
denoted
by
such terms
as
happy,
on-edge,
de-
jected,
and
yearning.
Second,
we
identify
a
category
of
Affec-
tive-Behavioral
Conditions
to
accommodate terms such
as
cheerful,
grouchy,
and
mournful
for
which
both
affect
and
be-
havior
are
focal.
Finally,
we
identify
a
category
of
Affective-
Cognitive
Conditions,
in
which
both
affect
and
cognition
are
focal,
as
they
are for
terms such
as
encouraged,
malice,
despair,
and
worried.
These three
categories
together constitute what
we
call
the
Affective
Conditions
class
and are
enclosed
in a box in
the
lower
left
of
Figure
1.
The two
remaining Mental
Conditions
are
those
for
which
affect
is not
focal
(although
it is, of
course,
implicated).
One of
these,
the
Cognitive
Conditions
category,
comprises
the
referents
of
words
for
which
only
cognition
is
focal
(e.g.,
certain,
prejudiced,
bewildered,
and
surprised)
and
the
other,
the
Cognitive-Behavioral
Conditions
category, con-
sists
of
conditions
for
which
both cognition
and
behavior
are
focal,
such
as
those referred
to by
words such
as
careful,
greedy,
and
virtuous.
These
two
categories,
shown
in the box in the
lower
right
of
Figure
1,
are
referred
to as the
Cognitive
Condi-
tions
class.
We
hypothesized that
the
best examples
of
emotion words
would
be
ones that
refer
to
internal
(as
opposed
to
external)
conditions, those that
refer
to
mental
(as
opposed
to
physical)
conditions,
and
those that
have
a
significant
focus
on
affect
in
the
sense just described.
If
this
is
accurate, then emotions
are
2
When
expressions
such
as
Internal
Conditions,
External
Condi-
tions,
and
Subjective
Evaluations
are
used
specifically
as
names
for hy-
pothesized
psychological
conditions,
they
will
be
differentiated
with
ini-
tial
capitalization
throughout
this
article.
1
There
is, of
course,
a
world
of
difference
between
words
and
their
referents.
In
general,
it is
clear
when
we are
talking
about
words
and
when
we are
talking
about
their
referents.
However,
on
occasion,
for
simplicity
of
exposition,
we use
phrases
such
as
"Subjective
Evaluation
Words"
as a
shorthand
for
technically
more
correct
phrases
as
"words
that
refer
to
members
of the
category
of
subjective
evaluations."
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AFFECTIVE
LEXICON
753
PHYSICAL*
BODILY
STATES
«.g.,
mrotiMd
AFFECT-
FOCAL
BEHAVIOR*
FOCAL.
COGNITION-
FOCAL
AFFECTIVE-
BEHAVIORAL
CONDITIONS
AFFECTIVE-
COGNmVE
CONDmONS
BEHAVIORAL-
COGNIT1VE
CONDITIONS
AFFECTIVE
STATES
COGNITIVE
CONOmONS
•.g.,
•nommuMM
Figure
I.
Taxonomy
of
psychological
conditions
proposed
by
Ortony,
Clore,
and
Foss (1987).
The
psycho-
logical
conditions
of
interest
are
shown
in rectangles, and the
features
that
differentiate
them
are
shown
in
ellipses.
all
members
of one of the
three
affect-focal
categories that con-
stitute
the
Affective
Conditions
class,
perhaps
with
the
more
prototypical
ones
being
located
in the
Affective
States category.
Although
in
particular
cases
the
decision
as to how
focal
affect,
behavior,
and
cognition
are is
sometimes
difficult
to
make,
the
reliability
between
us
was
high
overall
Thus,
to
summarize,
we
have
proposed eight
different
catego-
ries that
can be
collapsed
into
four
broad classes.
The
categories
are
Subjective Evaluations
and
Objective Descriptions
(which
together
constitute
the
External
Conditions
class),
Cognitive
Conditions
and
Cognitive-Behavioral
Conditions (the Cogni-
tive
Conditions class), Physical
and
Bodily
States
(which
is a
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

754
G.
CLORE,
A.
OKTONY,
AND M.
FOSS
class
of its
own),
and the
categories
of
Affective
States,
Affec-
tive-Cognitive
Conditions,
and
Affective-Behavioral Condi-
tions (which constitute
the
Affective
Conditions
class). This tax-
onomy
emerged
from
an
analysis
of a
sample
of
almost
600
English
words (see Appendix
A) in
which
we
considered
to
what
kind
of
condition each term referred.
As
already
indicated,
a
more detailed account
of the
linguistic
and
psychological
bases
for
these distinctions
can be
found
in
Ortony
et
al.
(1987).
Overview
The
main purpose
of
this
study
was to
test
the
validity
of the
classification
scheme just outlined.
The
scheme
is
emphatically
not
a
theory
of
emotion
but is an
approach
to
identifying
those
psychological
states
that need
to be
accounted
for in
theories
of
emotion.
We
attempted
to
determine whether
the
categories
could
be
discriminated
by
using
ratings
of the
degree
to
which
words
presented
in two
different
linguistic contexts were judged
to
refer
to
emotions.
The two
contexts
(feeling
something
vs.
being something)
were
intended
to
maximize sensitivity
to the
difference
between genuine emotions
and
nonemotional
states,
but we
expected
combinations
of
these
ratings
not
only
to
sepa-
rate emotion
from
nonemotion
terms
but
also
to
discriminate
the
other psychological conditions
in the
taxonomy
as
well.
Our
underlying
assumption
in
this study
was
that
feeling
something
is
much more
likely
to
seem
like
an
emotion
than
is
being
something. Consider,
for
example,
the
word
ignored.
We
assumed that subjects would
be
more
likely
to
indicate
that
feel-
ing
ignored
refers
to an
emotion more than
being
ignored
does.
If
being
ignored
is not
judged
to be an
emotion,
we are not
will-
ing
to
accept
the
word
ignored
as
referring
to an
emotion.
This
is
not to
deny
that
feeling
ignored
is
an
emotional
feeling;
it can
be. It is
likely
to
involve
such emotions
as
disappointment
and
hurt
feelings.
However,
the
word
ignored,
by
itself,
does
not re-
fer
to an
emotion;
it is
entirely possible
for
someone
to be ig-
nored (and even
to be
aware
of it)
without
caring
about being
ignored and, hence, without experiencing
any
concomitant
emotion.
Saying
"I
feel
ignored"
then,
is an
elliptical
way of
saying
something
similar
to "I am
being ignored,
and I
care that
I
am
being
ignored."
The
emotional content
of
feeling
ignored,
therefore,
derives directly
from
the
feeling
part
rather than
from
ignored
itself.
Consequently, being ignored need
not
involve
anything
emotional, although believing that
one is
ignored
may
well
be the
cause
of an
emotion (Ortony,
in
press).
The
fact
that
the
word
feeling
tends
to
bestow emotional
meaning
on
terms that accompany
it
could
constitute
a
prob-
lem
for
studies
in
which words
are
presented
without
a
context,
as is
usually
the
case.
Consider again
the
example
of the
word
ignored.
If one
were
to
present subjects with
the
word
ignored
in
isolation
and ask
them
to
rate their confidence
that
it
referred
to an
emotion, some subjects might interpret
it in
terms
of the
nonemotional
fact
of
being
ignored,
whereas other subjects
would
probably interpret
it in
terms
of the
emotional
feelings
implicit
in
the
expression./ee/mg
ignored.
The
result would then
be an
apparent lack
of
agreement about whether
or not
ignored
referred
to an
emotion.
This
problem
is
particularly
acute
with
Objective
Descriptions such
as
ignored,
abandoned,
and
abused,
which
we
have elsewhere (Ortony
&
Clore,
1981)
re-
ferred
to as
other
action words.
To
reduce
the
prospect
of
such
interpretational
ambiguity, subjects
in
this study were
first
asked
to
consider
the
form
that
we
thought
was
more
likely
to
be
judged
emotional
(i.e.,
the
feeling
form).
It was
assumed that
the
elicitation
of
feeling
judgments
along with
being
judgments
would,
by
sensitizing subjects
to the
difference
between
the two
forms,
reduce
the
possibility
of
subjects spontaneously reinter-
preting
the
being
form
as the
feeling
form,
thus reducing
the
likelihood
of
erroneous ratings
of
words
in the
being
form.
Syntactic
Forms
Although
most
of the
words used
as
stimuli were adjectives
or
adjectival
forms
(past
participles),
some emotion words
in
English
are
better,
or
only,
lexicalized
as
nouns
or
verbs. There-
fore,
we
used
three
lists
in
this study:
an
adjective list,
a
noun
list,
and a
verb
list.
Words
from
the
adjective list were presented
in
both
the
feeling
and
being forms. However, nouns
and
verbs
do not
lend themselves
to the
same feel-be manipulation
and
were
thus
treated
slightly
differently.
Nouns
appeared
both
in
the
feeling
form
and as the
unqualified
noun because
the
feel-
ing-noun
form
and the
noun-alone
forms
can,
for
practical
purposes,
be
treated
as
equivalent
to the
feeling-adjective
and
the
being-adjective
forms,
respectively.
Verbs
appeared
in
each
of
three
forms.
For
example,
the
verb
despise
appeared
as
feel-
ing
despised,
as
being
despised,
and as
despising
(someone).
Adding
the
present (active) participle
form
(e.g.,
despising
(someone))
is
important
for
detecting emotions
because
in
some
cases
the
past
participle
cannot
refer
to
Internal Conditions
in
the
context
of
being (e.g.,
being
despised),
whereas
the
present
participle
can.
This
turns
out to be
generally true
of
noncaus-
ative
verbs. Consider,
for
example, admire
and
hate. Neither
being
admired
nor
being hated
refer
to
Internal Conditions and,
hence, cannot
be
candidates
for
emotions,
but
admiring
(some-
one)
and
hating
(someone)
do
refer
to
Internal Conditions
and
are
thus
potential
candidates.
Although much
has
been written
in
linguistics about
the na-
ture
of
causative verbs
(e.g.,
Shibatani,
1976),
in the
context
of
this study,
one can
determine
the
difference
between
a
causative
and a
noncausative
verb
by
considering
what
is
implied about
the
grammatical subject
and
object
(in the
active
voice).
For
noncausative verbs,
the
focus
is on
some
state
or
activity
of the
grammatical subject.
For
example,
if
we
say
"John
hates Mary,"
John (the subject
of the
verb)
is the
experiencer
of the
affective
state.
On the
other
hand,
in
causative verbs
the
focus
is on
some
resultant
state
of the
person
who is the
grammatical object
of
the
verb.
If we say
"John
irritates
Mary," Mary (the object)
is
the
experiencer
of
that
affective
state.
In
other words, there
is a
syntactic
constraint
that precludes
the
possibility
of the
present
participle
of
causative verbs
(e.g.,
irritating)
from
being candi-
dates
as
emotions.
This constraint does
not
apply,
however,
to
the
present
participles
of
noncausatives
(e.g.,
hating)
or to the
past
participles
of
causatives
(e.g.,
irritated).
Hypotheses
As
a
test
of the
proposed
structure
for
the
affective
lexicon,
we
examined
the
discriminability
of the
various
categories
using
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AFFECTIVE
LEXICON
755
ratings
of the
words
in the
different
linguistic contexts.
A
con-
sideration
of the
ways
in
which
the
ratings
of
words
in
each
category
should
be
similar
and
different
in
these
two
linguistic
contexts
led to the
following
predictions:
1.
Words that
refer
to
Affect-Focal
Mental Conditions
(i.e.,
the
three
categories
in the
Affective
Conditions class) should
show
relatively
little
difference
between
the
being
form
and the
feeling
form,
both tending
to be
rated
as
emotions.
For
exam-
jAe,
feeling
angry
and
being
angry
should both
be
rated
as
emo-
tions.
The
basis
of
this prediction
is the
assumption that
the
function
of
feeling
is to
indicate that
an
emotion
is
experienced
in
association
with
the
predicated situation.
If,
however,
the
predicated situation
is
itself
an
emotion
(e.g.,
being
angry),
then
the
feeling
form
may
intensify
the
focus
on the
emotional
na-
ture
of the
state
but it
cannot create
it,
because
the
emotion
is
already presupposed
in
virtue
of the
presence
of the
emotion
word.
2.
Words
that
refer
to
Objective Descriptions
or
Subjective
Evaluations
(i.e.,
in the
External
Conditions class)
should
be
rated
as
referring
to
emotions
in
their
feeling
form
to a
signifi-
cantly
greater degree than
in
their
being
form.
For
example,
feeling
abandoned
and
feeling
stupid
are
likely
to be
rated
as
more
emotional than
are
being
abandoned
and
being
stupid.
This
is
because
in
their
feeling
forms,
the
emotional
aspects
of
expressions associated
with
being
in the
predicated situation
are
highlighted. Whereas
in
their
feeling
forms
these terms
can
be
expected
to
receive ratings similar
to
those
for
emotion
terms,
in
their being
forms
they
should
be
quite
discriminable
from
emotions, because they share none
of the
critical
features
of
emotions
(which
include being
an
internal, mental state
with
a
focus
on
affect).
We
predicted,
therefore,
that
the
ratings
of
terms
that
refer
to
Objective Descriptions
and
Subjective
Evalu-
ations
would
tend
to
have
large
differences
between
the
feeling
and
being
forms.
3.
Subjects should indicate
only
marginal
confidence
that
words
referring
to
Cognitive
or
Cognitive-Behaviora)
Condi-
tions
(i.e.,
in the
Cognitive Conditions class)
refer
to
emotions,
regardless
of the
form
being
rated.
We
expected these words
to
be
judged
as
somewhat emotional because they possess
many
of the
features
of
emotions
(indeed,
all of the
features
except
a
focus
on
aflect).
Furthermore, because
the
possession
of a
focus
on
affect
is a
matter
of
degree, terms
classified
as
Cognitive
and
Cognitive-Behavioral
Conditions
may be the
least discrimina-
ble
from
clear cases
of
emotion
terms.
4.
Words
that denote Physical
and
Bodily
States should
show
little
difference
between
the two
forms,
with subjects
showing
little
or no
confidence
that
they
refer
to
emotions
in
either
form.
For
example,
we
expected that
mi&x!
feeling
hungry
nor
being
hungry
would
be
rated
as
referring
to
emotions. These states
possess
few
of
the
features
of
emotions
in
that they
are not
even
good
examples
of
mental conditions.
5.
Additionally,
a
specific
prediction
pertaining
only
to
verbs
was
that
when
rated
as
emotions,
only
the
present (active) parti-
ciples
of
noncausative
verbs could
be rated as
emotions,
not
their
past
participles.
For
example,
admiring
(someone)
might
be
rated
as an
emotion
but
being
admired
could not. Con-
versely,
causative
verbs
could
be
rated
as
emotions
in
their past
participle (passive)
forms
but not in
their present participle (ac-
tive)
forms.
Thus,
being
frightened
might
be
rated
as an
emo-
tion
but
frightening
(someone)
would not.
In
summary,
with
respect
to how
emotional they seem
in the
contexts
of
feeling
and
being
(i.e.,
how
confident
subjects
are
that
the
phrases
refer
to
emotions),
we
predicted that
the
Affec-
tive
Conditions
would
be
high
on
both
forms,
that Cognitive
Conditions
would
be
moderate
on
both,
that
Physical
and
Bod-
ily
States
would
be low on
both,
and
that External Conditions
would
be
high
on the
feeling
form
but low on the
being
form.
The
ratings
for
verbs,
however,
should also depend
on
their
caus-
ative-noncausative
nature and,
concomitantly,
on
whether they
appear
as
present
or
past participles.
Method
In
the
study,
435
undergraduate psychology students served
as
sub-
jects.
They participated
in
groups
of 20 to 30,
with each person rating
a
subsample
of
the
pool
of
affective
terms.
The
study
was
designed
so
that
the
average rating
for
each item
would
be
based
on
approximately
20
observations,
although
the
actual range
was
from
18
to 80.
The
stimuli (see Appendix
A)
were
basically
the
same
wards
used
to
construct
the
taxonomy
of
psychological conditions described
in Or-
ion
y
et
al.
(1987)
except
that,
in
some
cases,
data
were
collected
on
additional syntactic
forms of the
same
word.
The
stimulus
set
included
most
of
the
words used
in
published
reports
of
studies
of
emotion (e.g.,
Bush,
1973;
Dahl&
Stengel,
1978,
Davitz,
1969;
Russell,
1980).
Because
the
predictions
involved
the
feel-be
distinction
and
because
adjectives
fit readily
into
the
feeling
and
being
contexts,
adjectives
and
verbs
in
adjectival (past
participle)
form
were
generally
chosen
as
stim-
uli in
preference
to
other
possible
syntactic
forms
when
the
meaning
was
judged
to
be
comparable.
The
basic
pool consisted
of
357
adjectives,
101
verbs,
and 20
nouns,
but the
noun
forms
of
53
of
the
adjectives
and
54
of the
verbs
were
also examined,
for a
total
of 585
words.
The
sublist
each
subject considered included
adjectives,
nouns,
and
verbs,
but no
subject
rated
the
same item
in
more than
one
syntactic
form.
As
already
indicated,
we
sought
to
prevent
subjects
from
reinterpret-
ing
items
in the
being context
as
though
they
had
been presented
in the
feeling
context.
For
this reason,
the two
contexts
were always presented
together,
with
the
feeling
form
immediately preceding
the
being
form,
in
the
belief that
having
subjects make successive judgments about
an
item
in its
different
contexts would
deter
them
from
reinterpreting
the
being
form
as
feeling.
For
example,
we
thought that
if
subjects
always
made
their
judgment about
being
abandoned
in
conjunction
with
their
judgment
about
feeling
abanttoned,
they would
be
unlikely
to
confuse
the
two.
For
adjectives,
ratings
were
collected
for
words
in the
feeling
context
(e.g.,
feeling
confused)
and
then
in the
being context
(e.g.,
being
confused).
For
verbs,
in
addition
to
these
forms
of the
past participle
(e.g.,
feeling
haled
and
being
hated),
subjects rated
the
present
partici-
ples (e.g.,
hating
(someone)).
Nouns,
on the
other
hand,
appeared
first
in
the
context
of
feeling
(e.g.,
feeling
hate)
and
then without
a
context
(e.g..
hate).
Subjects
were
instructed
to
indicate their confidence that
what
was
being
referred to by
each
of
the
phrases
in
which
the
words
occurred
(e.g.,
feeling
alone)
was an
emotion. Ratings
were
made
on a
4-point
scale anchored
at
one
end by the
words
certain
it
isn
'I
and
at
the
other
by the
words
certain
it is.
Intermediate scale points
were
labeled
suspect
it
isn't
and
suspect
it is.
Results
The
main
purpose
of the
study
was to
assess
the
soundness
of
our a
priori
or
rationally
derived
classification
system.
This
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

756
G.
CLORE,
A.
ORTONY,
AND M.
FOSS
involved
first
testing
the
discriminating power
of the
contrast-
ing
linguistic
contexts
of
feeling
and
being.
The a
priori
catego-
ries
were subjected
to a
discriminant analysis
to
discover
if
there
existed linear composites
of the
feeling
and
being ratings that
significantly
discriminate between
categories.
A
second task
was
to
determine
whether
or not any
such linear combinations
conformed
to
predictions.
Verbs
Before
analyzing
the
data
as a
whole,
we
examined
the
ratings
of
the
verbs alone.
The
main hypotheses
of the
study were
framed
in
terms
of a
contrast between
feeling
x and
being
x
where
x is an
adjective
or
past participle.
In the
case
of
verbs,
however,
the
predictions
for
past (passive) participles
only
apply
to
those
of
causative verbs, such
as
annoyed
and
frightened,
not
to
those
of
noncausatives,
such
as
admired
and
hated.
For
non-
causative
verbs,
we
argued (see Hypothesis
5)
that
only
present
(active)
participles (such
as
admiring
or
hating)
could
be
rated
as
emotions, whereas this
is
impossible
for the
present partici-
ples
of
causative verbs
(such
as
annoying
and
frightening).
The
test
of
these hypotheses involved
101
verbs,
21
of
which
were
noncausative.
The
results show that
of the 22
verbs
with
the
highest rating
in the
present
participle
form,
18
were non-
causatives.
Of
these
18,
17
appeared
originally
in one of the
affect-focal
groups
in the
taxonomy (the
affect-focal
groups
are
perceived
to
contain
the
best examples
of
emotions). These
17
verbs
are
admire,
appreciate,
desire,
enjoy,
grieve,
hate,
love,
resent,
adore,
despise,
detest,
disapprove-of,
dislike,
forgive,
like,
loathe,
and
want.
Only
4 (of the first 22)
were causative
verbs
(cheer,
inspire,
terrify,
and
excite).
Consistent
with
expec-
tations,
the
distribution
of the
ratings
for
present
participles
was
virtually
dichotomized, with
the
noncausatives dominating
the
upper
tail.
The
clear-cut
nature
of
this
result
suggests
the
cor-
rectness
of our
analysis
of
this
as a
logical,
as
opposed
to a
psy-
chological,
consideration.
Eight-Category
Discriminant
Analysis
The
ratings
on
feeling
and
being (averaged
across
subjects)
for
the 564
words (the total pool minus
the
21
noncausative
verbs)
were then submitted
to a
canonical discriminant analysis
by
using
the
CANDISC
program
in SAS
(SAS Institute,
1985).
The
analysis
was
intended
to
determine
how
well
the
eight
cate-
gories could
be
discriminated
on the
basis
of
subjects' ratings
of
the
terms
in
their
feeling
and
being
forms.
The
eight
catego-
ries
were
Affective
States,
Affective-Cognitive
Conditions,
Affective-Behavioral
Conditions, Cognitive-Behavioral Condi-
tions,
Cognitive Conditions, Physical
and
Bodily
States,
Subjec-
tive
Evaluations,
and
Objective
Descriptions.4
The
analysis yielded
two
significant
discriminant functions.
The
adjusted
canonical correlations were sizable
for
both
dis-
criminant
functions:
Rs = .75 and
.59,
F(14,1110)
=
70.22
and
^6,
556)
=
50.93;
o>mll/,,2
= .56 and
.16.
The
ow,,2
statistic
is
based
on
Tatsuoka
(1970).
The
uw,,2
value
for the
second
discriminant
function
is the
proportion
of
variance explained
after
the
variance related
to the first
function
is
removed.
The
first
function
represents
the
confidence
that terms
refer
to
emo-
tions.
It was a
composite
of the
being
and
feeling
forms
(stan-
dardized
coefficients
are
1.07
and
0.50, respectively)
and
pre-
dicted
scores
from
this
function
correlated
highly
with both
be-
ing
and
feeling
(r = .98 and
.91,
respectively).
The
second
dimension
was the
difference
between
the
feeling
and
being
forms
(2.08
and
—
1.93,
respectively).
It
represented
what
was
unique
to the
feeling
ratings
after
controlling
for the
being rat-
ings
(the part correlation with
the
second discriminant
function
was
.98).
The finding of two
significant discriminant
functions
confirms
that
the
manipulation
of
context
did
indeed induce
different
criteria, resulting
in
additional
items being judged
as
emotions
when
presented
in the
feeling
as
opposed
to the
being
context.
To
determine
how
well
these
two
variables discriminated
the
criterion
categories,
we
computed
the
Mahalanobis
distances
among
the
centroids
of the
categories.
The
distances
were
sig-
nificantly
different
for all but 3 of the 28
possible
pairs
of the
eight
criterion categories
(p <
.001,
see
Kshirsagar.
1972,
p.
146).
The
three
that
were
not
discriminable
were
not at all
sur-
prising.
The two
categories
constituting External Conditions,
namely
Objective
Descriptions
and
Subjective
Evaluations,
were
not
distinguishable
from
one
another. Also,
Affective-Be-
havioral
Conditions were
not
distinguishable
from
either pure
Affective
States
or from
Affective-Cognitive
Conditions.
No a
priori predictions were made about
the
discriminability
of
these
particular
categories.
The
canonical discriminant analysis
found
that
the
eight cat-
egories were distinguishable
on the
feel-be ratings.
To
examine
the
degree
of
overlap
of the
distributions
and to see
where
in
the
feel-be space each word
lies,
we
conducted
a
classificatory
discriminant analysis (see Tatsuoka, 1970)
by
using
the
DIS-
CRIM
program
in SAS
(SAS Institute,
1985).
This analysis indi-
cated
for
each
word
the
category whose
centroid
was
nearest
(in
terms
of
Mahalanobis distance
and
assuming equal
a
priori
probability
of
membership).
The
overlap
in the
distributions
was
as
expected. Categories with shared
properties
tended
to
overlap.
For
example,
of
those
Affective
State words
not
associ-
ated with (i.e., nearest
to)
their
own
centroid,
78%
were associ-
ated with
one of the
other
Affective
Conditions
(Affective-Cog-
nitions
or
Affective-Behaviors). Cognitive
Condition
words
were
next
most likely
(after
their
own
centroid)
to be
associated
with
Affective-Cognitive
Conditions,
followed
by
Physical
and
Bodily
States. Objective Descriptions
and
Subjective Evalua-
tions
were
also
found
to be
overlapped
highly
(over half
of the
words
not
associated
with their
own
category were
associated
with
the
other).
Eight-Category
Cluster
Analysis
Another
way to see the
organization
of the
categories
is to
examine
the
distances between
the
centroids. Therefore,
we
*
The
canonical
discriminant
analysis
procedure
assumes
equal
co-
variances
in the
groups
to be
discriminated.
Yet
this
was not the
case,
because
the
correlation
was
notably
lai^er
for the
Physical
and
Bodily
States
than
for the
others.
Although
the
procedure
is
somewhat
robust
against
this
violation,
the
results
should
be
considered
with
this
fact
in
mind.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AFFECTIVE
LEXICON
757
conducted
a
(complete-linked)
hierarchical cluster
analysis
on
the
Mahalanobis
distances among
the
centroids
of the
eight
cat-
egories.
Four clusters were clearly delineated (see Figure
2). One
might
be
called
the
Physical
and
Bodily
States
cluster,
as it
con-
sisted
only
of the
Physical
and
Bodily State words
and was
iso-
lated
from
the
other
clusters.
Another might
be
referred
to as
the
External
Conditions cluster;
it
consisted
of the
Objective
Descriptions
and the
Subjective Evaluations. This
was the first
cluster
to
form,
consistent
with
the finding
that
the
distance
be-
tween
their
centroids
was not
significantly
different.
A
Cogni-
tive
Conditions cluster
was the
next
to
form;
it
consisted
of the
Cognitive-Behavioral
and the
Cognitive categories. Finally,
an
Affective
Conditions cluster emerged,
consisting
of the
Affec-
tive,
the
Affective-Behavioral,
and the
Affective-Cognitive
cate-
gories.
These
four
clusters
correspond
clearly
to the
four
main
classes
of
conditions
in the
highlighted (boldfaced) boxes
in
Fig-
ure
1.
It
is
interesting
to
compare
the
clustering solution
with
the
kinds
of
differences
predicted. Because
their
ratings were
ex-
pected
to
differ
only
in
degree,
the
Affective
Condition words
(for
which
both
feel
and be
ratings
should
be
high)
should
be
most
similar
to the
Cognitive Condition words
(for
which
both
ratings
should
be
moderate). They should
be
less
similar
to the
External
Condition words
(high
feel
ratings
and low be
ratings)
because
of the
expected discrepancy
on the be
ratings
for the
two
classes.
Finally,
they
should
be
least similar
to the
Physical
and
Bodily
State words because
they
were
expected
to
differ
on
both ratings. This pattern
of
similarity
is
supported
by the
tree
structure shown
in
Figure
2. The
Cognitive cluster
is the
closest
to the
Affective
cluster
and
lies between
it and the
External
clus-
ter.
The
Body
States
cluster
is
distant
from
everything
but
most
similar
to the
External cluster.
There
is not
much
discriminability
within
the
Affective,
Cog-
nitive,
or
External
clusters. That
is, in the
Affective
cluster
as a
whole,
the
Affective,
Affective-Cognitive,
and
Affective-Behav-
ioral categories
all
looked alike.
The
same
was
true
within
the
Cognitive
cluster
for the
Cognitive
and the
Cognitive-Behav-
ioral categories
and
within
the
External cluster
for
the
Objective
Description
and the
Subjective Evaluation
categories.
None
of
the
predictions
concerned
the
discriminability
of
these
catego-
ries.
Therefore,
subsequent analyses
were
directed
at
distin-
guishing
the
words
as
classified
more broadly into these more
inclusive
but
more distinct
classes,
namely
the
Affective,
Cogni-
tive,
Bodily,
and
External Conditions
classes.
Four-Class
Discriminant
Analysis
The
same discriminant analyses that were
conducted
on the
eight categories
were
performed
on the
four
broader
classes.
The
analysis yielded essentially
the
same
two
discriminant
functions:
Rs = .72 and
.57,
F(6,
1118)
=
138.79
and
F(2,
560)
=
131.50,
amull?
=
-51
and
.16. Again,
we
analyzed
the
Mahalanobis
distances among
the
centroids
and
found
that
all
were
significantly
different
from
each
other
(p <
.001).
Figure
3
shows
the
extent
to
which
the
four
major
classes
were
discrim-
inable
by the
limited
information used
in
this
analysis.
The fig-
ure
shows
four
isodensity
contours,
one for
each class. Each
contour represents
a
region
within
which
68% of the
distribu-
tion
falls
(one standard deviation around
the
centroid,
assum-
ing
bivariate normality).
As
in the
case
of the
eight
categories,
a
classificatory
discrimi-
nant
analysis
was
also
applied
to the
four-class
data.
The
per-
centage
of
words associated
with
their
own
class
was 76% for
the
Affective
Condition words,
81%
for the
Body State
words,
71%
for the
External
Condition words,
and 49% for the
Cogni-
tive
Condition words.
The
patterns
of
association
among
the
classes
can be
understood readily
by
referring
to
Appendix
B,
which
shows
the
assignment
of
words
to
each
of
the
four
classes.
Whereas Appendix
B
shows
how
terms
from
each
of the
four
rationally derived
classes
were
classified
on the
basis
of the
data,
another
way
of
examining
the fit
between
the
predicted
and
ob-
tained
classifications
is to
consider
the
correspondence
between
the
rationally derived classification
and the
empirically derived
classification
as
shown
in
Table
1.
The
main diagonal reveals
that
in
all
cases
the
majority
of
items were assigned
to
their
a
priori
classes.
It
also shows that
the
most common
misclassin-
cation
was of
noncognitive items into
the
Cognitive Conditions
class. This
is
also evident,
of
course,
from
Figure
3,
which
shows
the
Cognitive Conditions class
as
having
the
most overlap with
the
others.
A
simple
test
of the first
four
hypotheses
we
described
in
the
introduction
of
this
article
can be
provided
by
comparing
the
sums
and
differences
of the
means
for the
feeling
and
being
rat-
ings.
Hypothesis
1
proposed that
for
emotion terms
(which,
we
claim,
are
predominantly members
of the
Affective
Conditions
class)
both
feeling
and
being ratings should
be
high,
so
that
there
should
be
relatively little
difference
between
the
two.
The
mean
ratings (represented
on a
0-1
scale)
of
the
items classified
preex-
perimentally
as
Affective
Conditions yielded
a
relatively
large
sum
(1.36)
and
small
difference
(.18),
which
when
compared
with
the
values
for the
other classes,
confirms
this prediction.
Hypothesis
2
predicted that
the
External Conditions class
would
have
a
large
difference,
with
high
feeling
ratings
and low
being ratings.
One
would
expect this
to
lead
to a
moderate sum.
The sum of
these ratings
was in
fact
.96 and the
difference
was
.37
(more than twice that
for the
Affective
Conditions class).
Hypothesis
3
maintained that
the
difference
in
ratings
for
items
in
the
Cognitive Conditions class
would
be
relatively small
and
that
the sum
would
be
only
moderate,
with
neither
very
high
feeling
nor
being
ratings.
This prediction
was
also
confirmed,
with
a
feeling
rating
of .60 and a
being rating
of
.37,
giving
a
sum of .97 and a
difference
of
.24. Finally,
for
Body State
terms,
Hypothesis
4
proposed that ratings would
be
low
on
both
forms,
giving
a
small
sum and a
small
difference.
The sum for
Body
State words
was .63 and the
difference
was.
16.
The
reliability
of
these
differences
among
the
four
classes
is
reflected,
of
course,
in the
results
of the
discriminant analysis just
reported,
which
snowed
that each class
was
significantly
different
from
all of the
others.
Discussion
To
distinguish
emotions
from
nonemotions,
Ortony
et
al.
(1987)
proposed
a
general structure
for the
affective
lexicon.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

758
G.
CLORE,
A.
ORTONY,
AND M.
FOSS
Affective
States
Affective
Cognitive
Conditions
Affective
Behavioral
Conditions
Behavioral
Cognitive
Conditions
Cognitive
Conditions
Objective
Descriptions
Subjective
Evaluations
Physical
&
Bodily
States
0.0
0.5
1.0
1.5
2.0
NORMALIZED
MAXIMUM
DISTANCE
Figure
2.
Complete
link
cluster
analysis computed
on the
Mahalanobis
distances
between
the
centroids
of
eight
categories
of
psychological
conditions.
The
structure
was
used
to
distinguish various kinds
of
psycho-
logical
conditions.
The
study
we
have
presented
here
attempted
to
assess
the
validity
of
that proposed structure.
Our
results
showed
that
the
four
broad classes
of
psychological states
and
conditions
distinguished
in the
proposed
taxonomy
were
empir-
ically
discriminate.
All
predictions about
the
patterns
of
rat-
ings
that
should
characterize
the
broad
classes
of
psychological
conditions
were
confirmed.
In
addition,
we
found
support
for
the
discriminability
of the
eight more
specific
categories
in the
eight-category
discriminant
analysis.
We had
expected that
the
best
examples
of
emotion terms
would
be
those
referring
to
conditions
that
are
states,
that
are
internal
as
opposed
to
exter-
nal,
that
are
mental
as
opposed
to
physical,
and
that
have
a
sig-
nificant
focus
on
affect
(i.e.,
the
class
of
Affective
Conditions).
The
data
clearly
support this position,
suggesting
that member-
ship
in the
Affective
Conditions class,
by
using
some sort
of em-
pirical
procedure such
as the one we
have
described, constitutes
a
theoretically
and
empirically defensible
criterion
for a
word
to
count
as
an
emotion term.
The
data
show
that
the
clearest cases
of
emotions
were
pro-
vided
by
terms
referring
to
conditions
we had
classified
in the
Affective
Conditions class
and the
clearest cases
of
nonemotions
were
those
we had
classified
as
Physical
and
Bodily States.
As
predicted,
these
had the
highest
and
lowest
emotion
ratings,
re-
spectively.
Neither
differed
much
with
respect
to
whether
feeling
or
being
forms
were
considered (Hypotheses
1 and 4). By
con-
trast,
terms
in the
External Conditions class
were
discriminable
by
being
rated
as
nonemotions
in the
being
form
but as
emo-
tions in the
feeling
form
(Hypothesis
2). We had
predicted that
terms
in the
Cognitive Conditions class
would
receive
only
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AFFECTIVE
LEXICON
759
1.
0
0.
9
0.
8
0.7
0.
6
F
e
e
1
0.5
i
n
g
0.4
0.3
0.
2
0. 1
0.0
A
P
C
E
:
Affective
:
Physical
:
Cogn
i t i ve
:
External
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Being
Figure
3.
Isodensity
contours
representing
68%
of
the
distribution
(one
standard
deviation)
for
each
of the
four
major
classes.
moderate ratings
in
both
forms
and
hence that they would
show
only
small
differences
in
their being
and
feeling
forms
(Hypoth-
esis
3).
This prediction
was
confirmed
in
that
the
Cognitive
Conditions terms were
significantly
different
from the
others
in
the
predicted direction. Finally,
with
regard
to
verbs,
we
showed
that terms referring
to
emotions were much more
likely
to be
found
among
the
present participle (active)
forms of
noncaus-
ative
verbs than
of
causative verbs (Hypothesis
5).
As
we
mentioned
at the
beginning
of
this article,
a
common
approach
to the
study
of
emotion
has
been
to use
scaling
meth-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

760
G.
CLORE,
A.
ORTONY,
AND M.
FOSS
Table
1
Frequencies
(and
Row
Percentages)
of
"Correct"
ami
"Incorrect"
Classifications
in the
Four
Classes
Rationally
derived
Empirically
derived
classification
classification
Affective
Cognitive
Body
External
Totals
Affective
234(76%)
37(12%)
6(2%)
30(10%)
307
Cognitive
19(14%)
66(49%)
23(17%)
26(19%)
134
Body
2(5%)
5(14%)
30(81%)
0(0%)
37
External
6(7%)
17(20%)
2(2%)
61(71%)
86
Totals
261
125
61
U7
564
Note.
"Correct"
classifications
are
denned
as
those
empirically
derived
classifications
that
match
the
rationally
derived
ones
(diagonal).
"Incor-
rect"
classifications
are
those
empirically
derived
classifications
that
do
not
match
(off-diagonal).
ods
to
determine
the
structure
of
emotions.
We
criticized such
attempts
to
investigate
the
structure
of
emotions
by
arguing
that
inadequate attention
has
been paid
to the
selection criteria
for
stimuli.
It is
interesting,
therefore,
to
consider
how the
stim-
uli
used
in
such
studies
would
be
classified
by
using
our
ap-
proach.
To
examine this issue,
we
examined
the
stimuli
from
two
frequently
cited
and
representative approaches
of
this kind
(Plutchik,
1980; Russell,
1980)
to
determine
to
which
classes
they
were
assigned
when
using
our
empirically based classifica-
tion
procedure.
Plutchik
(1980)
indicated that
40
words
were
used
in a
study
he
conducted.
Of
these
40
words,
34
were also
used
in our
list,
11
of
which
failed
to
satisfy
our
empirical crite-
rion
for
emotion words.
Of the
remaining
six
(agreeable,
dis-
trustful,
inquisitive,
intolerant,
puzzled,
and
receptive),
we are
confident
that none
would
have
been
classified
as
referring
to
affect-focal
mental states. This means that granted this
as-
sumption,
at
best
17 of
Plutchik's
40
words
are
questionable
examples
of
emotion words.
A
more pessimistic interpretation
would
be
that nearly
half
of the
words
do not
refer
to
emotions
at
all.
Similarly,
of the 28
words used
by
Russell
in a
number
of
studies
(e.g.,
Russell,
1980),
5
(i.e., nearly 20%)
fell
into
one
or
other
of our
nonaffective
categories. There
is, of
course,
no
presumption
here that
the hit
rate
of the
feel-be criterion
is
perfect.
One
might
easily make
a
case that some
of the
terms
our
empirically based procedure
has
classified
as
nonemotions
do in
fact
have
a
significant
focus
on
affect,
in
which case they
ought
to be
classified
as
affective
(i.e., emotional)
states
and
vice
versa.
At the
same time,
few
of the
words empirically
classified
as
nonemotions appear
to be
good examples
of
emotion terms.
Another
kind
of
comparison between
our
results
and
those
obtained
in
other studies
is
possible.
As an
example,
we
shall
consider
three
of the
stimuli used
in the
study
reported
by
Plut-
chik
(1980),
in
which
the
results
were
presented
as
evidence
for
a
circular model
of the
similarity structure
of
emotions. Now,
to the
degree that such
a
structure really
does
represent
the
sim-
ilarity
structure
of
emotions,
it
follows
that
the
closer
two
items
are
to one
another,
the
more
likely
they
should
be to be
judged
similar
on any
dimension that genuinely
characterizes
their
ref-
erents.
It
seems
to us
that
the
degree
to
which
a
word
is
judged
as
referring
to an
emotion
is
an
example
of one
such dimension;
consequently,
items that
are
highly
similar
ought
to
receive
comparable ratings
on
this dimension. Indeed, Plutchik's
data
show
fury
as
being
relatively
close
to
anger,
and our
procedure
classified
both
of
these words
as
emotions.
This
is as it
should
be.
However, Plutchik's data
also
show
defiant
as
being close
to
anger,
even
closer
in
fact,
than
Jury
is to
anger.
We
believe
that
in
theory
it
ought
to
follow
from
this that these
two
items should
be
even
more
likely
judged
as
highly
similar
on
other relevant
dimensions.
However,
they
are
classified quite
differently
both
in
our a
priori classification
and in our
empirically based one.
Specifically,
whereas
anger
is
classified
as an
emotion,
defiant
is
not:
It is
classified
as a
Cognitive-Behavioral condition.
But if
subjects
have
their doubts about
the
status
of
defiant
as an
emo-
tion,
how can it be so
near
(i.e.,
similar)
to
anger
in
Plutchik's
spatial representation?
We
would speculate
that
fury
and
anger
really
are
similar
in
most
respects
except intensity
and are so
judged.
However,
defiant
and
anger
are
not. Accordingly, judg-
ments about
the
similarity between them
(or
between
defiant
and any
other "genuine" emotion) might
be
based
on a
different
criterion: Perhaps
they
are
judged similar because
they
tend
to
co-occur.
If all the
stimuli
had
been
emotions,
the
likelihood
of
such
criterion
shifts
in
judgments would have been reduced
significantly.
This example
is
intended
to
underscore
the
main
motivation
for our
insistence
on
attending
to the
question
of
the
selection
criteria
for
stimuli
in
emotion studies: There
have
to
be
some,
and
neither other people's stimuli
nor
dictionaries
and
thesauri constitute
scientifically
adequate criteria
for
selecting
emotion stimuli, even though
they
may
constitute
the
most
usual
ones.
One
other interesting aspect
of the
data
is
that
they
suggest
that people
have
a
difficult
time distinguishing between intense
cognitive
states
and
affective
states.
Thus,
for
example,
aston-
ished,
bewildered,
flabbergasted,
and
amaz«/were
all
classified
preexperimentally
as
referring
to
Cognitive Conditions
but
sub-
jects'
ratings
of
them were more
similar
to
those
for
Affective
Conditions. Taking these
data
at
face
value suggests that other-
wise
nonemotional
(e.g., Cognitive
or
Body)
states
may
seem
more
emotional
as
they become more intense. Alternatively,
confusions
between intense cognitive
states
and
emotions
may
arise
simply
because intense cognitive
states
are
likely
to
cause
emotions.
For
example, being bewildered
is
likely
to
lead
to
frustration
and
distress, even though
the
phrase itself
refers
merely
to a
cognitive
state
of
confusion.
Within
the
emotion
literature,
we
know
of no
prior attempt
to
answer
the
questions
posed
here, although Shields
(1984)
raises
the
same question about
the
status
of
terms used
in
emo-
tion
research.
By
using
a
sample
of 60
candidate
terms,
she
demonstrated that subjects judged
many
of
them
not to be
emo-
tions. However,
she did not
seek
to
impose some
sort
of
struc-
ture
on the
affective
lexicon
as a
whole,
and she
used
a
relatively
small
sample
of
words. Most research
on
emotion words
has
been concerned with
characterizing
the
dimensional
structure
of
emotions,
often
as the
basis
for
proposing
a
theory
of
emo-
tion.
As
discussed previously, neither
of
these goals
character-
izes
our
research.
Rather,
we
have sought
to
differentiate
the
major
kinds
of
psychological states
and
conditions
to
which
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AFFECTIVE
LEXICON
761
terms
in the
affective
lexicon
refer.
These conditions include
not
only
emotional
states
but
also cognitive
states,
bodily
states,
and
so
forth.
Nevertheless,
there
is
some overlap
in the
categories
we
have
used
and
those
proposed
for
other purposes
by, for ex-
ample,
Allport
and
Odbert
(1936)
and
Norman
(1967).
These
authors
were
primarily interested
in
separating terms referring
to
personality
traits
from
other
person-descriptive
terms rather
than
in
separating emotion
from
nonemotion
terms.
It is of in-
terest,
however,
to
note that
they
also
found
it
necessary
to
sepa-
rate Subjective Evaluations such
as
wonderful,
contemptible,
and
weird
from
terms with more descriptive
content.
In
addi-
tion, they also worried about
differentiating
states
from
non-
states,
primarily
because they were interested
in one
kind
of
nonstate,
that
is,
traits (whereas
we
were interested
in one
kind
of
state,
i.e.,
emotions).
In our
previously
proposed
classifica-
tion
(Ortony
et
al.,
1987),
terms
were
also coded
with
respect
to
their
stateness.
When
the
distribution
of
states
and
nonstates
was
examined,
it
appeared that there were
no
psychological
states
in
either
the
External Conditions
or the
Cognitive-Behav-
ioral Conditions
and
that there
were
few
if any
nonstates
in the
Affective
Conditions. Hence,
as in
Norman's
(1967)
system,
our
system
separated
states
from
nonstates.
Allen
and
Potkay
(1981),
however,
have
been critical
of at-
tempts
to
distinguish
states
from
traits. They
focus
in
part
on
the
difficulty
of
distinguishing
traits
from
states
when
a
single
term
(e.g.,
happy
or
proud)
can
refer
to
either. They
see
evidence
in
terms such
as
happy
and
proud
that states
and
traits
are
fuzzy
sets that resist clear distinction.
In our
view,
the
difficulty
of
cataloging terms such
as
happy
and
proud
is not
that
the
state
and
trait
categories
have
fuzzy
boundaries
but
simply
that
there
are two
distinct meanings represented
by
each
of
these words.
In
ordinary discourse,
for
example,
one
must make clear
whether
one is
talking
about someone being happy
in the
mo-
ment
(the state
reading)
or
happy
in
general (the trait reading).
This
need
to
"disambiguate"
the
state
from
the
trait meaning,
however,
is not
evidence
for the
fuzziness
of the
state-trait
boundary.
It may or may not
turn
out to be
useful
to
think
of
the
state-trait
boundary
as
fuzzy,
but the
fact
that people
some-
times
use the
same terms
in
their assertions about momentary
states
as
they
do in
their assertions about enduring traits
is not
evidence
for
that
position.
Our
results
show
that
the two
sets
of
ratings were quite
infor-
mative.
However,
inspection
of the
cases
of
words
associated
with
a
centroid
other than their
own
(predicted one) suggests
various
possible
difficulties
with
this approach.
One
such
prob-
lem is the
tendency discussed earlier
for
ordinary cognitive
states
to
seem more emotional
as
they become more intense.
Another complication
can be
seen
in the
case
of
such
cognitive
terms
as
conceited,
serious,
and
foolish.
Subjects
were
in-
structed
always
to
consider
the
more emotional reading when
terms
could
be
interpreted
in
more than
one
way,
but
they
ap-
parently
did not
always
do so. For
example, these terms were
all
associated
with
the
External Conditions centroid.
We had
expected that these terms would show
the
pattern
for
Cognitive
Conditions
by
receiving moderate ratings
in
both
forms.
How-
ever,
whereas subjects
rated
them
as
emotions
in the
feeling
forms,
these words received
low
ratings
in
their
being
forms
and, consequently exhibited
the
External
Conditions pattern.
This
may
have
been
a
result
of
subjects interpreting
construc-
tions
such
as
being
serious
as
ways
of
behaving rather than
as
the
psychological
states
of
mind that
we had
intended.
Finally, many
of the
cases
in
which
a
word
is
associated
with
another category result
from
the
simple
fact
that
the
boundaries
of
these categories
are
such
that there
are
necessarily
many
bor-
derline
cases. Thus,
for
example,
such words
as
determined,
dis-
illusioned,
and
lively
were
classified
preexperimentally
as
Cog-
nitive
Conditions
but
were
associated
in the
data
with
the
Affective
Conditions
class.
Each
of
these terms
could
be
said
to
have
some
affective
focus; their classification ultimately
rests
on
the
degree
to
which
the
affective
aspect
is
seen
as
focal
in the
meaning
of the
word,
a
decision that
in the end
remains
a
judg-
ment
call.
Conclusion
The
belief underlying this project
is
that developing
an
ade-
quate theory requires that
the set of
phenomena
to be
explained
be
specified
clearly.
In the
area
of
emotion this
is
rarely
at-
tempted. Accordingly,
our
study
is
part
of an
attempt
to
develop
criteria
for
isolating terms that
refer
to
emotions
from
the
rest
of the
affective
lexicon.
To do
this,
we first
carefully
analyzed
the
referents
of a
large
pool
of
candidate terms.
The
outcome
of
that endeavor
was a
taxonomy
of
Affective,
Cognitive, Bodily,
and
External conditions.
We
then
assessed
the
results
of
this
project
by
attempting
to
classify
the
same terms into
the
same
taxonomy
on the
basis
of
empirical
data.
In
the
past,
scaling studies
have
been
a
common
but not al-
ways
satisfactory approach
to the
problem
of
identifying
emo-
tion
terms.
The
difficulty
with
simply
asking subjects
to
rate
whether
a
particular term
is a
good example
of an
emotion
is
that
the
linguistic context
in
which
subjects
implicitly consider
the
terms
is
usually
uncontrolled
and
unknown.
We
suspect
that
poor
examples
may find
their
way
into published lists
of
emotion terms because
they
are
considered,
at
least implicitly,
in
the
context
of
feeling
rather than being.
We
noted that many
distinctly poor examples
of
emotions
seem
quite
emotional
when
considered
in the
context
of
feeling
something
as
opposed
to
being something. Therefore,
in the
study
reported
here,
sub-
jects
were
asked
to
rate
each word
in
both contexts.
We ex-
pected emotions
to
appear
as
equally good examples
in
both
forms
and
nonemotions
to be
rated
as
poor
examples
in one or
both
forms.
Although
it
would
have
been
too
optimistic
to ex-
pect that these
two
judgments
by
themselves
would
reproduce
the
richer,
eight-category
taxonomy,
we did
expect
the
ratings
to
distinguish between
the
four
broad
classes
that
we
believe
constitute
the
underlying referential structure
of the
affective
lexicon.
Our
results showed that
the
pattern
of
ratings
in the
contrasting
linguistic
contexts
did a
remarkably good job. How-
ever,
we
should emphasize that
although
we
have
used
feeling
and
being ratings
as a
convenient
way of
testing
our
proposals
concerning
the
underlying structure
of the
affective
lexicon,
we
do not for a
moment believe that
the
feel-be distinction itself
plays
any
role
in
determining
that
structure.
The
fact
that
different
kinds
of
affective
conditions
can be
distinguished
in a
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

762
G.
CLORE,
A.
ORTONY,
AND M.
FOSS
feel-be
space does
not
establish that
the
corresponding dimen-
sions
determine
the
psychological
structure
of the
domain.
In-
deed, this
was
part
of our
criticism
of
traditional scaling
ap-
proaches
to
emotions.
Rather,
we
believe that
the
psychological
foundations
of the
affective
lexicon
are
components
such
as in-
ternal-external,
mental-nonmental,
and
affect-behavior-cog-
nition,
in
terms
of
which
the
different
types
of
affective
condi-
tions
are
distinguished.
References
Abelson,
R. P.
(1983). Whatever became
of
consistency
theory?
Person-
ality
and
Social
Psychology
Bulletin,
9,
37-54.
Allen,
B. P., &
Potkay,
C.
(1981).
On the
arbitrary
distinction
between
states
and
traits.
Journal
of
Personality
and
Social
Psychology,
41,
916-928.
Allport,
G. W., &
Odbert,
H. S.
(1936). Trait names:
A
psycho-lexical
study.
Psychological
Monographs,
47
(\,
Whole
No.
211).
Averill,
J. R.
(1975).
A
semantic
atlas
of
emotion
concepts.
JSAS
Cata-
log
of
Selected
Documents
in
Psychology,
5,
330. (Ms.
No.
421)
Bush,
L. E.
(1973). Individual differences multidimensional scaling
of
adjectives
denoting
feelings.
Journal
of
Personality
and
Social
Psy-
chology.
35,
50-57.
Carroll,
J. B.
(1959).
[Review
of The
measurement
of
meaning}.
Lan-
guage.
35,
58-77.
Dahl.
H.,
&
Stengel,
B.
(1978).
A
classification
of
emotion words:
A
modification
and
partial
test
of de
Rivera's
decision
theory
of
emo-
tions.
Psychoanalysis
and
Contemporary
Thought,
1,261-312.
Davitz,
J. R.
(1969).
The
language
of
emotion.
New
York:
Academic
Press.
Kshirsagar,
A. M.
(1972).
Multivariate
analysis.
New
York:
Marcel
Dekker.
Morgan,
R.
L.,
&
Heise,
D. R.
(1986).
The
structure
of
subjective
emo-
tions.
Unpublished manuscript, Indiana University.
Norman,
W. T.
(1967).
2800
personality
trait
descriptors:
Normative
operating
characteristics
for a
university
population. Unpublished
manuscript,
University
of
Michigan, Department
of
Psychology.
Ortony,
A. (in
press).
Is
guilt
an
emotion? Cognition
and
Emotion.
Ortony,
A.,
&
Clore,
G. L.
(1981,
August).
Disentangling
the
affective
lexicon.
Paper presented
at the
third annual conference
of the
Cogni-
tive
Science
Society,
Berkeley,
CA.
Ortony,
A.,
Gore,
G.
L.,
&
Foss,
M. A.
(1987).
The
referential structure
of the
affective
lexicon.
Cognitive
Science
11,
361-384.
Osgood,
C.
E.,
May,
W.
H.,
&
Miron,
M. S.
(1975).
Cross-cultural
uni-
versals
of
affective
meaning.
Urbana:
University
of
Illinois
Press.
Osgood,
C.
E.,
Suci,
G.
J.,
&
Tannenbaum,
P. H.
(1957).
The
measure-
ment
of
meaning.
Urbana:
University
of
Illinois
Press.
Russell,
J. A.
(1980).
A
circumplex
model
of
affect.
Journal
of
Personal-
ity
and
Social
Psychology,
39,
1161-1178.
Plutchik,
R.
(1962).
The
emotions:
Facts,
theories,
and a new
model.
New
York:
Random House.
Plutchik,
R.
(1980).
A
genera]
psychoevolutionary
theory
of
emotion.
In
R.
Plutchik
& H.
Kellerman
(Eds.),
Emotion:
Theory,
research,
and
experience.
Vol.
1.
Theories
of
emotion (pp.
3-31).
New
York:
Academic
Press.
SAS
Institute.
SASuser's
guide:
Statistics,
19SS
edition.
Cary,
NC: SAS
Institute.
Shibatani,
M.
(Ed.).
(1976).
Syntax
and
semantics.
Volume
6: The
grammar
of
causative
constructions.
New
York:
Academic
Press.
Shields,
S.
A.
(1984).
Distinguishing between emotion
and
nonemotion:
Judgments
about
experience.
Motivation
and
Emotion,
8,
355-369.
Tatsuoka,
M. M.
(1970). Discriminant
analysis:
The
study
of
group
differences.
Champaign,
IL:
Institute
for
Personality
and
Ability Test-
ing.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AFFECTIVE
LEXICON
Appendix
A
Stimulus Words From
the
Affective
Lexicon
763
abandoned
abused
accept*
admiration
admire*
adore*
adventurous
affection
affectionate
afraid
aggravated
aggressive
aggrieved
agitated
agitation
agony
alarm
alarmed
alert
alertness
alone
aloof
amazed
amused
amusement
anger
angry
anguish
anguished
annoyed
antagonistic
anxiety
anxious
apathetic
apathy
apologetic
appreciate*
appreciation
apprehensive
approve-of*
argumentative
arousal
aroused
arrogant
ashamed
astonished
at-ease
at-peace
attracted
attraction
attractive
aversion
aware
awe
awestruck
awful
bad
baffled
beaten
beloved
benevolent
bereft
bewildered
bitchy
bitter
blue
bold
bored
boredom
brave
breathless
brokenhearted
burdened
calm
carefree
careful
careless
cautious
certain
charitable
charmed
cheated
cheered
cheerful
cheerfulness
cheerless
comfortable
(physically)
comfortable
(psychologically)
compassion
compassionate
competent
competitive
complacent
conceit
conceited
concern
concerned
confidence
confident
confused
confusion
conscientious
consolation
consoled
contempt
contemptible
contemptuous
contented
contentment
contrite
convinced
cooperative
courage
courageous
cowardly
crabby
crazy
critical
cruel
cruelty
crushed
curiousity
curious
cynical
daring
dazed
defeated
defensive
defiant
deflated
degraded
dejected
dejection
delight
delighted
dependent
(physically)
dependent (psychologically)
depressed
depression
deprived
desire
desire*
despair
desperate
despicable
despise*
despondent
determination
determined
detest*
devoted
devotion
disagreeable
disappointed
disappointment
disapprove-of*
discontented
discouraged
disenchanted
disgraced
disgust
disgusted
disheartened
disillusioned
dislike*
dismay
dismayed
displeased
dissatisfied
distress
distressed
disturbed
dizzy
dominated
doubt
doubtful
downhearted
dread
dreadful
dreary
droopy
drowsy
dull
eager
earnest
ecstasy
ecstatic
elated
elation
embarrassed
embarrassment
emotional
empathy
encouraged
encouragement
energetic
enjoy*
enjoyment
enthusiasm
enthusiastic
envious
envy
euphoria
euphoric
exasperated
exasperation
excited
excitement
exhausted
exhaustion
expectant
faint
faithful
fascinated
fascination
fatigue
fatigued
fear
fearful
fed-up
feverish
fine
flabbergasted
fond
fondness
foolish
forgive*
friendliness
friendly
fright
frightened
frustrated
frustration
fulfilled
funny
furious
fury
gaiety
generous
gentle
glad
gleeful
gloomy
glorious
glum
good
grateful
gratified
gratitude
greed
greedy
grief
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

764
G.
CLORE,
A.
OKTONY,
AND M.
FOSS
Appendix
A
(continued)
grief-stricken
grieve*
grouchy
guilt
guiltless
guilty
happiness
happy
hate
hate'
hateful
hazy
heart-stricken
heartbroken
heartened
heartsick
heartsore
heavy-hearted
helpless
hesitant
high
homesick
hope
hopeful
hopeless
hopelessness
horrible
horrified
horror
hostile
hostility
humble
humiliated
humiliation
hung-up
hunger
hungry
hurl
ignored
ill
ill-at-ease
impatience
impatient
impotent
impressed
in-love
inadequate
incensed
incredulous
indifference
indifferent
indignant
ineffective
infatuated
infatuation
inferior
inhibited
insecure
inspiration
inspired
insulted
interest
interested
intimate
intimidated
irate
irked
irritable
irritated
irritation
isolated
itchy
jealous
jealousy
jittery
joy
joyful
joyless
joyous
jubilant
kind
lazy
lighthearted
like*
lively
livid
loathe*
loneliness
lonely
lonesome
longing
lost (psychologically)
lousy
lovable
love
love*
lovesick
loving
low
lucky
lust
mad
malice
malicious
marvelous
meek
melancholy
merry
mischievous
miserable
misery
mistreated
mixed-up
modest
modesty
mortified
mournful
moved
nauseous
neglected
nervous
nervousness
nonchalant
nostalgia
nostalgic
numb
obstinate
odd
offended
on-edge
oppressed
optimism
optimistic
outrage
outraged
overconfident
ovejoyed
overwhelmed
pain
pained
panic
passion
passionate
pathetic
patient
peaceful
peculiar
peeved
perplexed
persecuted
pessimism
pessimistic
petrified
petty
phony
pining
pissed-ofT
pitiful
pity
placid
playful
pleasant
pleased
pleasure
powerful
prejudiced
pride
protective
proud
purposeful
quiet
rage
reassurance
reassured
rebellious
reckless
refreshed
regret
relaxed (physically)
relaxed (psychologically)
relief
relieved
remorse
repentant
resent*
resentful
resentment
resigned
respect
respect*
rested
restless
reverence
revived
ridiculous
rigid
rotten
sad
sadness
safe
sarcastic
satisfaction
satisfied
scared
scorn
scornful
secure
self-centered
self-confident
self-conscious
self-destructive
self-pity
self-satisfaction
self-satisfied
selfish
sensitive
(considerate)
sensitive
(easily
hurt)
sentimental
serene
serenity
serious
sexy
shaken
shame
shock
shocked
shook-up
shy
sick
sick-at-heatt
sickened
silly
sincere
skeptical
sleepiness
sleepy
slighted
sluggish
smug
solemn
solemnity
soothed
sore
(psychologically)
sorrow
sorry
spiteful
startled
strange
strong
(psychologically)
stubborn
stunned
(psychologically)
stupid
submissive
successful
suffering
superior
sure
surprise
surprised
suspense
suspicion
suspicious
sympathetic
sympathy
tender
tense
terrible
terrific
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AFFECTIVE
LEXICON
765
Appendix
A
(continued)
terrified
terror
thankful
thirst
thirsty
threatened
thrilled
thwarted
timid
tingly
tired
tolerant
tormented
touched
(psychologically)
triumphant
troubled
trust
trust*
trustworthy
unattractive
uncared-for
uncertain
uncomfortable
(physically)
uncomfortable
(psychologically)
uncooperative
uneasy
unfaithful
unfriendly
unfulfilled
unhappy
unimportant
uninterested
unlovable
unpleasant
unprotected
untroubled
untrustworthy
unworried
upset
uptight
useless
vain
vanity
vengeful
vigor
vigorous
violent
virtue
virtuous
vulnerable
want*
warm
warmhearted
weak
(psychologically)
weary
weird
welcome
well
willful
woe-stricken
wonder
wonderful
worried
worry
yearning
Note.
The
21
items marked with
an
asterisk
are the
noncausative
verbs.
Appendix
B
Comparison
of
Empirically
to
Rationally
Derived
Classifications
of
Affective
Words
The 564
words (i.e.,
the 585
words indicated
in
Appendix
A
less
the
21
noncausative verbs marked there
with
an
asterisk,
which
were ana-
lyzed
separately)
are
listed
first
according
to
the
four
(preexperimental)
rationally
derived
classes.
Within
each class
the
words
are
further
sorted
under
the
class
for
which
the
Mahalanobis
distance
from
the
word
to
the
centroid
of
that class
was
smallest.
In
other words,
the first
major
group
lists
all
those items that were
preexperimentally
classified
as
Affective
Conditions
and
shows
the
class
to
which
each
was
assigned
on
the
basis
of the
data.
Note that
the
Affective
Conditions class combines
the
Affective,
Affective-Behavioral,
and
Affective-Cognitive
categories,
that
the
Cognitive Conditions
class
combines
the
Cognitive
and
Cogni-
tive-Behavioral categories,
and
that
the
External Conditions class com-
bines
the
Subjective Evaluation
and
Objective Description categories.
Affective
Conditions
The
following
307
words
were
classified
preexperimentally
as
Affec-
tive
Conditions. Their average ratings
on
feeling
and
being
were
.774
and
.590, respectively.
The 234
words
(76.2%)
empirically classified
as
Affective
Conditions
(.809,
.648). Admiration,
affection,
affectionate,
afraid,
aggravated,
ag-
grieved,
agitated, agitation,
agony,
alarm, alarmed, amused,
anger,
an-
gry,
anguish,
anguished,
annoyed,
anxiety,
anxious, apathy, apprecia-
tion, ashamed,
at-peace,
attracted, awe, awestruck, bitchy,
bitter,
blue,
brokenhearted, cheered,
cheerful,
cheerfulness, comfortable
(psycho-
logically),
compassion, compassionate, concern, concerned, consoled,
contempt,
contemptuous, contented, contentment, crushed, dejection,
delight,
delighted, depressed,
depression,
desire,
despair,
desperate,
de-
spondent, disappointed, disappointment,
discontented,
discouraged,
disenchanted, disgust, disgusted, disheartened, dismay, dismayed,
dis-
pleased,
dissatisfied,
distress,
distressed,
disturbed, downhearted,
dread, ecstasy,
ecstatic,
elated, elation,
embarrassed,
embarrassment,
emotional, empathy,
enjoyment,
enthusiasm, envious,
envy,
euphoria,
euphoric, exasperated, exasperation, excited, excitement,
fear,
fed-up,
fond,
fondness,
fright,
frightened, frustrated, frustration,
fulfilled,
furi-
ous,
fury,
gaiety,
gleeful,
gloomy, glum,
grateful,
gratified,
grief-stricken,
grief,
grouchy, guilt, happiness, happy, hate, heart-stricken,
heartbro-
ken,
heartened, heartsick, homesick, hope,
hopelessness,
horrified,
hor-
ror,
hostile, hostility, humiliation, hurt,
ill-at-ease,
impatient,
in-love,
incensed, infatuated, infatuation, intimate, intimidated, irate, irked,
ir-
ritated, irritation,
jealous,
jealousy, joy,
joyful,
joyous, jubilant, kind,
lighthearted,
livid,
loneliness,
lonely,
lonesome,
longing,
love,
lovesick,
loving,
low,
lust,
mad, malice, melancholy,
merry,
miserable, misery,
mortified,
mournful,
moved, nervous, nervousness, optimism,
optimis-
tic,
outrage,
outraged, overjoyed, overwhelmed,
panic,
passion,
passion-
ate, peeved,
petrified,
pining,
pissed-off,
pity,
placid,
pleased,
pleasure,
pride, proud, rage,
regret,
relaxed (psychologically),
relief,
relieved,
re-
morse,
resentful,
resentment, sad, sadness,
satisfied,
scared,
scorn,
self-
conscious,
self-pity,
self-satisfaction,
self-satisfied,
sensitive (easily
hurt),
sentimental, shame, shock, shocked,
sick-at-heart,
solemnity, sor-
row,
sorry,
spiteful,
suffering,
suspense, sympathetic,
sympathy,
tender,
terrified,
terror, threatened, thrilled, tormented, troubled, uncomfort-
able (psychologically), uneasy, unhappy, upset, uptight,
vengeful,
warm-
hearted,
woe-stricken,
worried,
worry,
and
yearning.
The 37
words
(12.0%)
empirically
classified
as
Cognitive Conditions
(.634, .410). Amusement, apathetic, apologetic, apprehensive,
at-ease,
attraction,
aversion, burdened, carefree, charmed, cheerless,
contrite,
cowardly,
deflated,
devoted,
eager,
encouraged,
enthusiastic,
gratitude,
humble,
indignant, nostalgic, pained, pessimistic,
repentant, reverence,
satisfaction,
scornful,
serenity,
shook-up,
shy, sickened,
smug,
soothed,
sore
(psychologically),
thankful,
and
wonder.
The 6
words (2.0%) empirically
classified
as
Physical
and
Bodily
States
(.481,
.364). Devotion, encouragement, impatience,
nostalgia,
pessimism,
and
respect.
The 30
words
(9.8%)
empirically
classified
as
External Conditions
(.737, .402). Benevolent, calm, consolation, crabby, dejected,
fearful,
glad,
heartsore,
heavy-hearted,
high,
humiliated, insecure, irritable,
joyless, malicious,
offended,
on-edge,
peaceful, reassurance, reassured,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

766
G.
CLORE,
A.
ORTONY,
AND M.
FOSS
secure, serene, shaken, solemn, tense, timid, touched (psychologically),
triumphant,
unfulfilled,
and
warm.
Cognitive
Conditions
The
following
134
words
were
classified
preexperimentally
as
Cogni-
tive
Conditions. Their average ratings
on
feeling
and
being
were
.603
and
.365, respectively.
The
19
words
(14.2%)
empirically classified
as
Affective
Conditions
(.745,
.550).
Amazed,
argumentative,
astonished,
bewildered,
bored,
courage, dependent (psychologically),
determination,
disillusioned,
flabbergasted,
friendliness,
hopeful,
lively,
mixed-up,
sensitive
(consid-
erate), surprise,
surprised,
suspicious,
and
violent.
The 66
words
(49.3%)
empirically
classified
as
Cognitive Conditions
(.587, .359).
Adventurous,
aloof,
baffled,
cautious,
certain, charitable,
complacent,
confidence,
confident, confused,
confusion,
conscientious,
critical,
cruelty, curiosity, cynical, defensive,
defiant,
determined,
doubtful,
energetic, expectant,
faithful,
fascinated, fascination, gener-
ous,
greed, greedy, hesitant, hung-up, impressed, incredulous,
indiffer-
ent,
inspiration,
inspired,
interested,
lazy,
mischievous, modest, non-
chalant,
obstinate,
overconfident,
patient,
perplexed,
petty,
playful,
pur-
poseful,
reckless,
resigned, rigid,
self-confident,
sincere,
startled,
stubborn, stunned (psychologically), submissive, sure, trust, uncertain,
uncooperative,
unfriendly,
vanity,
vigcn;
vigorous,
virtuous,
and
willful.
The 23
words
(17.2%)
empirically classified
as
Physical
and
Bodily
States
(.450,
.295). Alert, alertness,
arrogant,
aware,
boredom,
careful,
conceit,
convinced,
cooperative,
curious,
doubt,
earnest,
hazy,
indiffer-
ence,
interest,
modesty, prejudiced, restless,
self-centered,
skeptical,
suspicion, tolerant,
and
virtue.
The 26
words
(19.4%)
empirically classified
as
External Conditions
(.675,
.304).
Aggressive,
antagonistic,
bold,
brave,
careless,
competitive,
conceited,
courageous, crazy,
cruel,
daring,
foolish,
friendly,
funny,
gen-
tle,
inhibited,
lost (psychologically), meek, protective, rebellious,
sarcas-
tic,
selfish,
serious,
silly,
stupid,
and
vain.
Physical
and
Bodily
States
The
following
37
words were classified
preexperimentally
as
Physical
and
Bodily
Conditions.
Their
average
ratings on
feeling
and
being
were
.393
and
.236,
respectively.
The 2
words
(5.4%)
empirically
classified
as
Affective
Conditions
(.736, .592). Arousal,
and
aroused.
The 5
words
(13.5%)
empirically
classified
as
Cognitive Conditions
(.584,
.405).
Jittery, refreshed, relaxed (physically), revived,
and
well.
The 30
words
(81.1%)
empirically classified
as
Physical
and
Bodily
States
(.338,
.184).
Breathless, comfortable (physically),
dazed,
dizzy,
droopy,
drowsy, exhausted, exhaustion,
faint,
fatigue,
fatigued,
feverish,
hunger,
hungry,
ill,
itchy,
nauseous,
numb,
pain, rested, sick, sleepiness,
sleepy,
sluggish,
thirst,
thirsty,
lingly,
tired,
uncomfortable (physically),
and
weary.
No
words (0.0%) were empirically classified
as
External
Conditions.
External
Conditions
The
following
86
words were
classified
preexperimentally
as
Objec-
tive
Descriptions.
Their average ratings
on
feeling
and
being
were
.666
and
.298, respectively.
The 6
words (7.0%) empirically
classified
as
Affective
Conditions
(.793, .529). Disgraced,
hateful,
hopeless,
isolated,
lovable,
and
slighted.
The
17
words
(19.8%)
empirically classified
as
Cognitive Conditions
(.580,
.304).
Beloved, competent, contemptible, dependent (physically),
disagreeable, dominated,
dreadful,
dreary,
dull,
peculiar,
powerful,
safe,
trustworthy,
untroubled,
unworried,
weak
(psychologically),
and
wel-
come.
The 2
words (2.3%) empirically classified
as
Physical
and
Bodily
States
(.455, .220). Phony,
and
uninterested.
The
61
words
(70.9%) empirically classified
as
External Conditions
(.684,
.276). Abandoned, abused, alone, attractive,
awful,
bad, beaten,
bereft,
cheated, defeated,
degraded,
deprived,
despicable,
fine,
glorious,
good,
guiltless,
guilty,
helpless, horrible,
ignored,
impotent,
inadequate,
ineffective,
inferior,
insulted, lousy, lucky,
marvelous,
mistreated,
ne-
glected,
odd,
oppressed,
pathetic,
persecuted,
pitiful,
pleasant,
quiet,
ri-
diculous,
rotten,
self-destructive,
sexy,
strange, strong
(psychologically),
successful,
superior,
terrible,
terrific,
thwarted, unattractive, uncared-
for,
unfaithful,
unimportant, unlovable, unpleasant,
unprotected,
un-
trustworthy,
useless, vulnerable, weird,
and
wonderful.
Received
June
19,1986
Revision received January
5,1987
Accepted
April
1,1987
•
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.