I

I NI

UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN

PRODUCTION NOTE
University of Illinois at
Urbana-Champaign Library
Large-scale Digitization Project, 2007.

T ol
39Os^ S~

-- 7Technical Report No. 293

QUANTITATIVE AND QUALITATIVE SOURCES OF AFFECT:
HOW UNEXPECTEDNESS AND VALENCE
RELATE TO PLEASANTNESS AND PREFERENCE
Asghar Iran-Nejad and Andrew Ortony
University of Illinois at Urbana-Champaign
October 1983

Center for the Study of Reading

READING
EDUCATION
REPORTS

6^.^

UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN

51 Gerty Drive
Champaign, Illinois 61820
B

^* '

"

BOLT BERANEK AND NEWMAN INC.

50 Moulton Street
Cambridge, Massachusetts 02238

The NationE
Institute c
Educatio
U.S. Department
Educati
Washington. D.C. 202

CENTER FOR THE STUDY OF READING

Technical Report No. 293

QUANTITATIVE AND QUALITATIVE SOURCES OF AFFECT:
HOW UNEXPECTEDNESS AND VALENCE
RELATE TO PLEASANTNESS AND PREFERENCE
Asghar Iran-Nejad and Andrew Ortony
University of Illinois at Urbana-Champaign
October 1983

University of Illinois
at Urbana-Champaign
51 Gerty Drive
Champaign, Illinois 61820

Bolt Beranek and Newman Inc.

50 Moulton Street
Cambridge, Massachusetts 02238

The research reported herein was supported in part by the National Institute
of Education under Contract No. HEW-NIE-C-400-76-0116, and in part by a
Spencer Fellowship awarded to the second author by the National Academy of
Education. We wish to thank Harry Blanchard, Jerry Clore, and Terry Turner
for their insightful comments on an earlier version of this paper.

EDITORIAL

BOARD

William Nagy
Editor
Harry Blanchard

Anne Hay

Wayne Blizzard

Patricia Herman

Nancy Bryant

Asghar Iran-Nejad

Pat Chrosniak

Margi Laff

Avon Crismore

Brian Nash

Linda Fielding

Theresa Rogers

Dan Foertsch

Behrooz Tavakoli

Meg Gallagher

Terry Turner

Beth Gudbrandsen

Paul Wilson

Valence Potential

Valence Potential

Abstract
Quantitative and Qualitative Sources of Affect:
affect is

Optimal-level theories maintain that the quality of

function

of

How

a
Unexpectedness and Valence Relate to Pleasantness and Preference

potential

arousal

quantitative

dimension.

An

view is that the

alternative

and

quantitative dimension merely modulates pre-existing qualitative properties

is

only

therefore

responsible for changes in the degree of affect.

In 1874, Wundt proposed the

quality of affect, whether it is positive or negative, has to be

separate
the

Thus, the

independent variable.

quantitative

dimension

was

treated

as

of

endings

in

inverted-U

intensity

and

hedonic

states.

moderate level is increasingly pleasant and beyond

manipulated

by

varying

the

degree

stories

that

were

were rated

story

stimulus

endings,

this

optimal-level

stimuli

increasingly

less pleasant up to some indifference baseline, from which

overall either positive or

Contrary to predictions of optimal-level theory, results showed

valenced

link

of

in

stimulus

intensity.

As

that
major

differently

to

According to Wundt, stimulus intensity up to a

point unpleasantness increases with increments

negative.

curve

In an experiment to compare these alternatives,

become
unexpectedness

classic

theoretical construct, the optimal-level curve stimulated a great deal of

judged the same on an expectation scale,

research,

especially

important

revision

during

the

1950's

when

the

hypothesis

underwent

an

very differently in hedonic tone and preference.

(McClelland, Atkinson, Clark, & Lowell, 1953; Haber, 1958).

According to the revised version, affect

physical

intensity

per

se,

but

of

is
the

"a

function,

not

of

size of the discrepancy between the

adaptation level (adaptation or expectation) of the organism, and

(perception)"

(Haber,

1958,

increasing

pp. 370).

Subsequent

major

the

developments

optimal-level theory were primarily due to Berlyne (e.g., 1960, 1973,

reintroduced

the

potential, i.e.,
defined

arousal

original

Wundt

curve,

replacing

intensity

arousal-inducing properties of external

potential

to

include

stimulus

9 74

in

a) who

with

arousal

stimulation.

Berlyne

not only intensity or discrepancy from

expectation but other "collative" variables

such

as

complexity,

incongruity,

conflict, and uncertainty.

The basic assumption underlying the optimal-level hypothesis

origin

of

the

quality

of

affect,

quantitative dimension (the quantity of

be

it

is

that

the

positive or negative, lies in the

arousal

potential,

the

size

of

the

Valence Potential

expectation,

from

discrepancy

etc.).

the quality of affect may be seen

origin of

in

from

quotation

following

the

italics

hedonic

"positive

increase

through a moderate
a

through

values

decrease

in

can

come

recent

optimal-level

literature,

researchers

(cf.,

quantity

Ortony,

1982;

the

valence

of

certainly, they never explore any other sources.

Gati & Tversky, 1982),

is more troublesome.
from

or

quantitative

the

treat

in

even

Furthermore,

(p. 8).

(e.g.,

It can
Gati

Stevens,

&

Tversky,

1957).

In

modulates

the proposal that it determines quality

be argued that quality

1982;
the

is

fundamentally

Iran-Nejad,
context

of

distinct

1980; Iran-Nejad &

affect,

the

qualitative/quantitative distinction is central, most notably, to the two-factor

theory of emotion proposed by Schachter and Singer (1962).

deal

origin

of

affect,

with the problem of testability often raised in

optimal-level

theory.

The

problem

arises

from

the

problem

and

support

below

the various stimuli presented to the subject

the

optimum,

then

an

representing

optimal-level.

sampling

of

Thus, Arkes

range

inverted-U must be found in order to

However, most optimal-level research does

optimal level a priori"

span

(p. 164).

not

specify

explained

the independent variable on only one side of

and

Garske

have

an

Consequently, empirical results

showing linear rather than curvilinear properties can always be

as

for instance,

as follows: "If an optimal level of a subject is known, and

the theory.

individual's

Arkes and Garske (1977),

concluded

that

"an

away
the

inverted-U

relation allows so many possible curves that the theory is difficult to refute."
In view of the amount of
and

continues

to

research that the inverted-U hypothesis

generate

(e.g.,

has

generated

Carrol, Zuckerman, & Vogel, 1982;

1967; Greenberg & O'Donnell, 1972; Karmel & Maisel, 1975; Zillmann,

Eysenck,

1980),

the

conclusion that the theory may be irrefutable and therefore "worthless" (Arkes &

While it is reasonable to assume that the quantitative dimension

intensity

stated

mechanism')

dimension as if it were the only source of affective quality (i.e.,

hedonic tone);

the

arousal when arousal has reached an uncomfortably high

in

level (the 'arousal-reduction mechanism')"

more

'arousal-boost

to

the

the complexities of

about in either of two ways, namely

(the

arousal

with

the

above

For instance, Berlyne (1974b)

causes affective quality.

that it

state

the sole source of the quality of affect, but they do

dimension is

quantitative

attempts

the optimal level (or point) is located.

added)

More recent statements of optimal level theory do not explicitly claim that

that

paper

the

difficulty of determining a priori where on the abscissa of the inverted-U curve

a theory
McClelland, Atkinson, Clark, and Lowell (1953) have developed
on Hebb's (1949) neurological model of the origin of
in part,
based,
to
According
level.
adaptation
affect and Helson's (1947) notion of
of affect,
origins
to the
as
discrepancy hypothesis
McClelland's
or
'positive affect is the result of smaller discrepancies of a sensory
adaptation level of the organism; negative
from the
event
perceptual
affect is the result of larger discrepancies' (McClelland, et al., 1953,

suggest

this

connection

(Haber, 1958, p. 370,

Valence Potential

In examining the qualitative/quantitative view of

clear statement of the quantitative

Haber (1958):

p. 43).

Garske, p. 164) is a disturbing one.

to draw such a conclusion.
is

One clear prediction that optimal-level theory makes

that a given degree of arousal potential cannot give rise

and negative affect.

of

However, we believe that it would be wrong

quality

from

to

both

positive

This may be contrasted with the hypothesis of independence

quantity,

and

the

corollary

hypothesis

that

qualitative variable that is the direct source (cause) of affect.

it

is

On this view,

under different qualitative conditions, a given degree of arousal potential

be

positive or negative.

the

can

In contrast to the notion of arousal potential, which

signifies, in part, the quantitative dimension,

we

refer

to

the

qualitative

Valence Potential

that a stimulus has for influencing

variable as valence potential: The potential

manipulation of
explanation

valence-specific biofunctional characteristics of the organism.

Valence Potential

the extent to which the subject has

of

an

appropriate

or

his bodily state, and (c) the creation of situations

from which

explanatory cognitions may be derived" (Schachter & Singer, 1962, p.

Two lines of research may be construed as having already

produced

results

One

line of

optimal-level
the

contradicting

inverted-U

quantitative

purely

hypothesis.

proper

382).

An

theorist can argue that the latter two manipulations, rather than

exerting their influence on the quality of affect in terms of cognitive labeling

research generally cited as

counter-evidence (see Walker, 1981) has involved the

operations,

use

of

gustatory

(e.g., Engle,

stimuli

1928;

Pfaffman, 1960,

1969).

do

so

in

terms

of

their

own arousal-inducing properties.

example, in the Schachter and Singer experiment, apart from the intended
stimuli (e.g.,

quinine) fail to result

in

at

pleasantness

1960,

1969)

and

others

their intensity (Engle, 1928),

regardless of
post-ingestion

intervene

factors

or

unpleasant

become

(see, e.g., Pfaffman, 1960).

have produced additional

arousal.

proper

their

that optimal-level theory can

survive

that

gustatory/sensory

Epinephrine-ignorant

explanation

for

such

findings.

arousal

symptoms,

as

result

of

subjective

uncertainty.

may

have

been

from

Engle's

have been additionally

Epinepherine-misinformed

aroused not only because of

curves

subjective uncertainty or

the discrepancy between

data

the

symptoms

they

were

such
expecting and those they actually experienced.

resulting

may

Furthermore, the affect-inducing

might "be brought together to form an

situation itself may have given rise to additional arousal, perhaps more so

inverted-U if

might deviate from the inverted-U curve

data

because they merely reflect peripheral activity and concluded that

those

lacking

He

injection but also because of

as

subjects,

Walker (1981),

subjects
reasoned

arousal through injection, other aspects of the experiment may

after

only

aroused
however, has argued

of

sugar) seem either to be pleasant

(e.g.,

direct

concentration

any

manipulation
(Pfaffman,

For

Certain

they were plotted against neural intensity

. .

. measured

at

for

an
the

anger

appropriate central site rather than in the sensory nerve" (p. 42-43).

than the euphoria condition as the Schachter and Singer data seem to

indicate.
The other line of

research, though not

commonly

discussed

in

connection
Similar problems arise in interpreting other experiments in this tradition.

with

the

inverted-U hypothesis, originated

in the work of Schachter and Singer

For

(1962).
from

the

Their theory of
quantity

Singer (1962)

of

emotion suggests a separation of

arousal.

However,

the

kind

of

neither the original Schachter and

experiment, nor studies adopting a similar attribution of

instance,

(e.g.,

White,

Fishbein,

&

Rutstein,

1981)

have

seconds

LA)

subjects then watched a

or

for

120

seconds

Arousal,

HA).

The

female

confederate

who

was

made

to

appear

(Low

Arousal,

either

highly

unequivocally
(High

Attraction,

HAT)

or

unattractive

(Low

Attraction,

LAT).

that the same degree of arousal can be both positive or negative--a

Subjective measures of

demonstration that we believe

be definitively refuted.

attraction

indicated

that

HA-HAT

liked

the

and HA-LAT subjects liked her less

than

to be necessary if the inverted-U hypothesis is to

confederate

The typical attribution of arousal

paradigm

more

involves
LA-LAT subjects.

"(a)

(High

arousal

attractive
demonstrated

had male subjects participate

in an exercise (the arousal manipulation) either for 15

videotape of
approach

White, Fishbein, & Rutstein (1981)

affect

the experimental manipulation of a state of physiological arousal, (b) the

than LA-HAT subjects,

The scores for LA-HAT, HA-HAT, LA-LAT,

and

subjects

HA-LAT

conditions

Valence Potential

inverted-U interpretation of

the data would only require the assumption that the

condition caused more arousal than the positive attraction

attraction

negative

An

inverted-U curve if they were arranged in that order.

an

constitute

would

arousal operations.

attribution of
does

cognition

generate

not

A second characteristic of our view is

affective

stimulus

that

is

independent

of

are

independent

of

quality-valence

meaning, and valence potential properties of the

This implies that cognition influences the quality and

their meaning potential.

condition.

Valence Potential

intensity of affect not directly but in terms of valence potential properties of

has

It is perhaps because of such considerations that optimal-level theory

the

stimulus

which,

presumably, exert their influence through certain affect-

continued to exert an influence on research in spite of evidence that appears to

specific areas of
in

can,
as

domains

Karmel & Maisel,

1975),

Zillmann, 1980),

&

and

prose

Russell,

comprehension

74

proposed by Zajonc (1980),

a),

(e.g.,

(e.g.,

appreciation

Brewer

&

the

of

cognitive content, as has been

in the same way that

(cold) cognition

to be considered as

of

attribution

a separate categorical

arousal

or

distinct

variable,

only

not

from

quantitative factors, but also from cognitive

other

Finally, the biofunctional

theory implies

that

factor

must

be

the

intensity

of

distinguished from the intensity of arousal dimension.

This contention is supported by evidence that cognition can

Our approach, which

directly

intensify

is
the experience of

affect without the mediation of autonomic arousal (Iran-Nejad,

1982), has,

biofunctional model of cognition (Iran-Nejad & Ortony,

in

quality of affect is distinct from the level of

arousal, they are markedly different in other respects.

on

occur

arousal
affect

based

can

1980; Moynihan & Mehrabian, 1981).

that the

maintain

absence

Empirically, this means that the quality of affect has

of affect.

absence

evaluation.

both

the

environmental

entertainment

media

1974),
and

19

1971,

Although the valence potential hypothesis and the

theory

in

occur

principle,

physiological
Lichtenstein, 1981; Kintsch,

affect

(e.g., Greenberg & O'Donnell, 1972;

aesthetics (e.g., Berlyne,

Mehrabian

(e.g.,

psychology

development

perceptual

infant

This notion assumes that the experience of

the brain.

diverse

The inverted-U hypothesis is employed in such

be inconsistent with it.

1983).
for present purposes, three important characteristics.

First,

it

claims

that
The quantitative factor manipulated in the present experiment is the

the

intensity

of
or

than quality somehow emerging from cognitive evaluations that label

unexpectedness.

It

must

be noted, however, that there exists no evidence

explain
that unexpectedness (or any other collative variable)

already-existing

level

exerts its influence on already-existing quality, rather

factor

arousal.

With

respect

to

the

empirical

exerts

its

quantitative

this

framework,

influence in terms of autonomic arousal even though many optimal-level theorists

assumption means that the initial valence potential

of

the

stimulus

must

be

had

an

assume that it
taken

into

For

account.

instance,

if

and

Schachter

Singer

had

influence
independent group of

subjects

rate

the

behavior

of

the

confederate

on

of

does.

Rather,

unexpectedness

we,
per

like
se

they,

examined

negative/positive

the situational cues in

those

scale,

on the experience of affect.

of autonomic

they would presumably have observed that

the anger condition were

initially

negative

and

(quantitative)
Since most

a
optimal-level studies do not involve direct measures

dichotomous

the

that

in the euphoria condition were initially positive, quite independently of

use of unexpectedness as a quantitative variable is entirely fair.

arousal,

the

Valence Potential

Our qualitative variable, the valence potential of the

distinguished

from

hedonic

tone.

Hedonic

tone

degrees of pleasantness or unpleasantness, while

stimulus

properties

stimulus,

must

refers to the experience of

valence

potential

that influence the sign of that experience.

refers

both the qualitative and quantitative aspects

affective

is

It

interaction between two independent

valence

potential

is

realized

continuous
dimensions,

as

affective

manifests itself in some quantity, that is,

experience.

Increasing the amount of

only

in more of

to a different

The use of
of

by analogy,

sugar) is independent of

results

or

of

whether

essential

of

from

Obviously,

experience,

in terms

the

resulting

causes.

In this sense, quality of affect and its

in much the same way that,

(e.g.,

variable

to

Hedonic tone,

therefore, is a combination of
functioning.

be

the
when

it always and only

degrees

of

affective

intensity are independent

nature

of

substance

(i.e., cannot arise from) the weight dimension.

such a substance from one arbitrary quantity to another

the same.

It does not

and cannot change the substance

substance (e.g., sugar to salt).

valence potential as

it

should

be

velocities

arise

from

levels of direction.
contribution

of

the

problem

as a dichotomous or trichotomous variable.

Valence Potential

contribution

of non-zero speeds to one of the two

The special case of zero

velocity

arises

speed to some third level of direction (i.e.,

but simply from zero speed.

"stationariness."

Notice,

Thus, it makes no
however,

that

sense

to

not

from

the

"no direction"),

talk

of

degrees

of

extremely low speeds may result in a

moving object seeming to be stationary (one cannot see the hour hand moving on a
clock

face).

But,

this

is

fundamentally

different

sense

of

the word

"stationary" from the genuine absence of velocity (and direction).

So too with affect:
states.

It

can

be

An organism

which is a continuous variable,

quantitative

factor.

is

of

valence

valence potential.

potential

be

in

function

levels

one

no

of

of

of

three

valence

hedonic

the quantitative

experiential
Hedonic tone,

potential

tone

dimension

arise
to

one

and

from

the

of

two

The special case of zero hedonic tone arises not

the quantitative

(i.e.,

quantitative dimension.

Non-zero

contribution of non-zero levels of

levels

can

in a positive, negative, or neutral state.

from the contribution of

an independent variable raises the

treated

10

valence),

dimension
but

to

simply

some

third

level

of

from a zero level on the

Thus, while positive and negative hedonic tone can vary

Although the phenomenological experience of affect can be positive, negative, or

in degree, it makes no sense to talk of degrees of neutrality.

neutral,

that extremely low levels on the quantitative dimension may result in a valenced

one

structure.
Since

cannot

assume

that

In fact, we propose that

valence

valence potential

straight

line.

forwards, it
a

has the same tripartite

has

no

neutral

level.

this is a crucial assumption, the following extended analogy will be used

to clarify it: Imagine an object able to move

is

potential

The

object

can

be

forwards

in one of

can be moving backwards, or it can be

continuous

or

backwards

three states.
stationary.

It

qualitative

can be moving

Velocity,

variable, is a function of direction and speed.

quantitative dimension, and direction is the

along

which

Speed is the

dimension.

stimulus seeming to be neutral.

the word

Notice, however,

But, this is a fundamentally different sense of

"neutrality" from the genuine absence

of

hedonic

tone

(and

valence

potential).

If valence potential is
attempts

to

determine

the

qualitative component of

the

indeed only a two-valued variable, it follows

valence

of

resulting

some

particular

hedonic

tone)

stimulus (i.e.,

must

avoid

that
the

mistaking

Non-zero
apparent neutrality (i.e.,

imperceptibly low levels of hedonic tone) for genuine

11

Valence Potential

minutely

are

Things that

neutrality (i.e., the absence of valence potential).

preference,

or

utility"

(Berlyne,

priori reason to equate hedonic tone and

evidence

suggesting

Moynihan & Mehrabian,

they

that
1981).

are

preference.

distinct

other, perhaps

because of

Indeed,

. degree

is

there

of

some

dimensions (see,

psychological

It seems entirely possible

additional

However, there is no a

1974b).

that

could

person

still preferring one over the

two objects as being equally pleasant while

judge

'hedonic tone' embraces

and, for example, preference: "the term

pleasure,

with

synonymous

were

Optimal-level theorists rarely distinguish between hedonic tone

tone.

hedonic

it

if

factors

qualitative

(cognitive)

level

optimal

This

contrast them with those of the valence potential hypothesis.

was done by manipulating the level of unexpectedness and

of

the critical conclusion information in story endings.

and then made

hedonic

tone and preference ratings.

to make the following predictions:

states

depending

information;
degrees

on

(because

the only source of affective quality),

the

pleasant

initial valence potential

and (b) identical

or

initial

and thus identical

unpleasant

hedonic

of the critical conclusion

levels of unexpectedness can result in

different

of preference, again depending on the valence potential of the critical

conclusion information.
assumed

manipulation

to

be

Furthermore, since

psychologically

hedonic

tone

and

preference

are

distinct, (c) there could be a contribution of

unexpectedness to preference after its effects through hedonic

tone

have

been

the

main

partialed out.

Method

the valence

potential

Subjects read stories

Sixty high school students

experiment.

The

majority

(grades

of

11 and

12)

participated

in

the subjects were female but the two sexes were

approximately evenly divided among the experimental conditions.

Design and Materials

would

Optimal-level theory

A 2 x 2 randomized factorial design was used with

two

levels

of

(a) the quality of hedonic tone (i.e.,

expectation

positive or negative) should be a direct result of

the

unexpectedness

of

(expected

vs

unexpected)

and

two

level

of

valence

potential

the
(positive vs negative).

critical

expectation

levels of unexpectedness can result in either

Design.
seem

is

Valence Potential

Subjects

The present experiment attempted to test the predictions of

and

potential

the

(e.g.,

interestingness).

theory

tone should be independent of

valence

positive or negative are, nevertheless, positive or negative.

So far we have been using the term affect as

12

Both factors were between-subjects.

conclusion information in story endings, and thus identical degrees of

unexpectedness should result in hedonic states of

unpleasant),

and

(b)

identical

identical degrees of preference.

significant

levels

of

the same quality (pleasant

unexpectedness

result in

In addition, (c) unexpectedness should make no

contribution to preference beyond its

contribution via hedonic

tone

contrast,

the

(because hedonic tone and preference are not distinguished).

predictions

should

or

of the valence potential hypothesis are: (a)

In

the quality of hedonic

The passages.
Thurmond

(1978).

The four passages

Each version consisted of

there were four stems and two endings.

Marilyn,

who

left

were

The

revised

versions

of

a stem and an ending.

basic

story

was

about

story

Altogether,
a

nurse,

the hospital where she worked after a late night shift.

she was driving home, she noticed

that

she

was

running

out

of

by

gas.

As
This

frightened her, especially because there had been a recent surge in muggings and

beatings in the area.

with

Gabriel,

birthday gift.

nice

unusually

she was slightly acquainted having been to his station for

He insists

awkward

She finds herself in an

The stem ends as

invitation.

the

and

situation

the

accepts

reluctantly

she follows him inside.

unexpected versions) or by signaling it

versions) in otherwise identical story versions.

in the stems (for

Note that,

as it

stems

the

from

the expected
is, the basic

is perhaps going to be raped/mugged and that Gabriel

is perhaps a rapist/mugger.

The column labeled "Critical Ending Information" in

Table 1 shows the gist of

the conclusion information

a (friendly) dog emerges from Marilyn's car, that she is

that

is

ending

for

negative and positive

was

information

This

safe, and that Gabriel is a Good Samaritan.

to

assumed

have overall positive valence potential.

The column marked

to

additions

the

"Critical

constructed

so

as

Stem

to

the

story

story

The expected and unexpected

main

thematic

were

versions

be semantically similar as much as possible, especially

their

begin

and

the

moment

the

with respect to the overall story content at

reading

the

shows

Information"

basic stem, as well as other important stem information, for

each of the four conditions.

that Marilyn

implies

stem

this

that she go inside his office to see

Unexpectedness was manipulated by withholding information

(of

an

He fills up the car and tells her that he has recently received

gas before.

it.

whom

station run by a person called

She decided to go to a gas

Valence Potential

14

Valence Potential

13

The main difference

ratings.

affective

finish

subjects

between the stem for the unexpected negative version and the basic stem was that
the

contained

former

sentence

noticed a police car behind her.

there

that while driving home Marilyn

indicating

It also contained information suggesting

"Marilyn discounted the few

be something suspicious about Gabriel:

might

disturbing rumors that accompanied his sudden

that

in

appearance

the

area."

This

Insert Table 1 About Here

stem,
In the ending for negative

story versions.
that

out

turns

Marilyn

her.

Thus, the critical conclusion information for this ending is

that

in

her

potential.
Marilyn
that
come,

Rather, the police

discover

heroin

is

this

information

was

assumed

to

Marilyn's

in

instrumental

In the ending for the positive valence potential

not

raped/mugged.

find

no rapist/mugger in the car.

hospital parking lot attendant.

valence

conditions,

again

Furthermore, Gabriel, who apparently suspected

"someone" was hiding in the back of Marilyn's car calls

but

negative

have

the

police.

Instead they find the dog of

Thus, the critical conclusion

information

They

implied

that

was

Gabriel

wolf

in sheep's clothing and

possibily a rapist/mugger and that Marilyn was perhaps going to be raped/mugged.

in
and

Overall,

arrest.

is not raped/mugged and that Gabriel is not a wolf

a drug dealer, and that Gabriel is an informer

Marilyn is

it

conditions,

potential

car

sheep's clothing.
arrest

valence

therefore,

The stem for the expected negative story contained

additional

information

that Marilyn was somehow involved with drugs, although the nature of

indicating

even

as

an

FBI undercover agent) was not

clear.

For example, the stem stated

that while driving "she looked forward to a long soak

while she soaked she planned

in

the

tub

stem

implied

that

Gabriel

danger of

for

probably a drug dealer.

being raped/mugged.

[and]

to skim through a new magazine and forget about her

involvement in the drug business." Thus, like the unexpected negative one,

the

or

nurse,

this involvement (e.g., as a pusher, in connection with her job as a

might

be

this

a rapist/mugger and that Marilyn was

However, this stem also implied that Marilyn

and that Gabriel was perhaps a police informer.

in
was

15

The unexpected positive version was constructed by

basic

the

thought she heard someone breathing behind her."

was

likely

again

to

inserted

this

stem,

while

be raped/mugged, Gabriel was not portrayed as a

to

Rather, the stem implied that he was

rapist/mugger.
intending

Information was also

Therfore, in

inside in an attempt to get her out of danger.

Marilyn

information

Gabriel probably saw someone hiding in the car and invited her

that

suggesting

adding

perhaps

Good

Samartan

Valence Potential

police car, she regrets having trusted Gabriel.

someone might be hiding in Marilyn's car--"she

that

implying

stem

16

Valence Potential

In the ending for the positive conditions, Gabriel locks the door,

car

It is clear now that both are safe.

door,

"a

They arrive and Marilyn and Gabriel go to the window

gun, and calls the police.

to watch.

gets

Then, when the police

open

the

large dog stepped out, obviously confused by the flashing lights

and sirens" and Marilyn realizes that it belongs

to the security

guard

of

the

hospital parking lot.

to save Marilyn from getting raped/mugged by a maniac probably hiding

Story characteristics.

The two story endings

were

constructed

priori

in the back of her car.

such
the

The expected positive stem contained information additional to that in

unexpected

stem--information compatible with the possibility that the

positive

dog of the hospital parking lot attendant was in the back

of

brief dialog the attendant tells Marilyn that his dog is lost,
bored

even

again,

possibility that
remained.

though
a

there was

This

ending

In

car.

that when it gets

back of my car," and that "he isn't there now."

"he goes and sleeps in the

However,

her

this

version

person

implied

that

in

was designated as "expected," the

the
Marilyn

back

of

Marilyn's

car

always

was again in danger of getting

that

the

critical conclusion information for the negative story versions

has negative valence potential and that

positive.

It

instance,

there

be rated as positive.
gets

her

just

First, in this

deserts

(i.e.,

one

for

the

negative

pair and one for the positive pair.

The ending for the

negative conditions begins with Gabriel turning quickly around and

door.

He

pulls a gun and

tells Marilyn that there is no birthday present.

this point, approaching squad

with flashing lights.
of

heroin.

locking

cars are heard and the gas station is soon

the
At

filled

Police officers search Marilyn's car, and find three bags

They come in, handcuff her and take her away.

As she

sits

in

the

versions

that

what

happened

potential

of

the

For

dealer,

Marilyn,

Secondly, contrary to what was

raped/mugged.

the arrest) was

rape/mugging).
critical

drug

In

might

think

not quite as bad as what

order

conclusion

One

to

confirm

information

of

that

the

the negative

the positive versions was positive, the

materials

Since affective judgments were to be made after reading the entire

story, steps were taken

stories.

guilty

gets arrested).

to Marilyn (i.e.,

versions was negative, and of

were normed.

is

is not a forgone conclusion that

ending, a

implied by the negative stems, Marilyn is not

valence

stems,

story

are at least two reasons why the negative ending might in fact

was expected to happen (i.e.,

four

positive

the endings are as they were intended to be in terms of valence potential.

loose in the area.

As, described earlier, two endings were associated with these

the

must be noted, however, that it

raped/mugged, if not by someone hiding in the car but by a rapist/mugger running

Gabriel was perhaps again going to be a Good Samaritan.

for

While

reading

to

avoid
a

story,

potential
a

reader

confounding
may

experience

in

norming
a

sequence of

alternating affective states. For the purposes of the

present

experiment,

critical

follows

the

state

is

the

last

one,

the

one

that

manipulation and determines affective judgments of

the subjects.

We

the

the

expectation
felt

that

17

the

valence of

Valence Potential

this state could be assumed to be determined only by the valence

potential of the critical conclusion information,

state

could

critical

and

the

intensity

of

this

conclusion

information.

In

order

to

take

the mental state that was assumed to be the basis of

judgements

and

experienced.
conclusion

account

that

of

all

other

of

this

had

their

response

(final)

affective

(prior) mental states subjects might have

In other words, although the valence

information

the

potential

of

the

critical

to be determined in the context of the rest of the

story, if subjects read a story and are then asked whether

negative,

might

it

is

positive

or

be based not just on the last affective state

that they experienced, but on some summary

judgment

earlier

of this potential contamination, two

states.

To

reduce

the

synopses were constructed, one

versions.

for

chance
the

positive

These synopses were constructed so as

of

and

the

one

sequence

for

have

immediately after reading them.

the

of

the

negative

to match as closely as possible

the representations of the semantic content of the stories that

presumably

negative) scale.

subjects

The following is

would

Valence Potential

Eight different judges rated the conclusion

for the negative versions and all of them rated it

be assumed to be determined by the degree of unexpectedness of the

consideration, the norming task had to distinguish between the valence potential
of

18

The dependent measures.

and

then stopped briefly to respond to two preending rating scales.

preending

preference

unexpected way, that is,

think

scale,

asked

would

subjects
like

to

"I

would

like

One

of

see

this

Subjects

pleasantness

rate

in

an

degree

to

choosing

ranging

from

recalling

the

this option, and none did, turned the page and

(extremely

to

rate

the

unpleasant)

ending

to 10

on

(extremely

pleasant), and on an expectation scale ranging from 1 (extremely unexpected)

to

10 (extremely expected).

the synopsis
The first scale

positive versions:
measured

Marilyn, a nurse, leaves the hospital where she works after a late night
shift.
A dog
belonging to
the
hospital parking
lot attendant is
sleeping in the back seat of her car.
She knows the dog
but she
does
not know that it is in the car. As she is driving, she notices that she
is low on gas.
She decides to go to a gas station whose
attendant
she
knows.
While
cleaning the back windows, the attendant sees something.
He thinks someone is probably hiding in the back of her car and gets her
out
by
inviting
her to go inside his office "to see the nice birthday
gift my sister gave me." Once inside, he calls
the police.
When the
police
come,
they find the dog. Marilyn notices that it is the dog of
the hospital parking lot attendant.

Eight adult judges rated the

ended);

end

the

In addition, there were five postending preference scales.

for the

scales,

in a way very different from the way the story makes me

They were then immediately asked

scale

the

to stop here and not read the ending." On the next page,

not

the ending.

Each scale

story

subjects were actually given the choice of reading the ending or

story.

stem

to indicate their degree of

it will end." The second scale asked the subject to

which

synopsis

Each subject in the main experiment read the

agreement with the statement: "I

read

the

as negative.

ranged from 1 (strongly disagree) to 10 (strongly agree).

the

of

seven

of

them

conclusion of

rated

it

as

this

positive

synopsis

(i.e.,

the

way

the

degree

of

agreement of

the subject with the statement,

"I would

like to read again a story with an ending of the same type (meaning expected
unexpected)

as

the

one

just

read."

postending preference scale, because it

for

the

type

of

ending

they

preference scales were indirect in

it

on a dichotomous (positive vs

had

We will refer to this as

asked subjects to rate their

just directly experienced.

that

they

asked

subjects

or

the direct
preference

The remaining
rate

their

preferences for types of story endings they had not actually encountered.

There

were four such scales.

The first (Scale 1) measured the

degree

to

to

which

the

subject "would like to read a passage with an unexpected ending if the ending is

19

the

words

expected/unpleasant,

and

pleasant." Scales 2, 3, and 4 repeated the same statement, replacing

with

unexpected/pleasant

20

Valence Potential

expected/pleasant,

Postending Ratings

first

Procedure

successful.

and

the

rating

scales.

the

they

do

not

like

to

read.

They

were

told that they would read two

the

the type of passage people

the

first passage was to determine

like to read.

Subjects then read the stem, responded

scales,

the

read

ending,

memory

this passage were

to

the

and

preending

that

rating

and responded to the postending rating scales.

response booklets were then collected and subjects

containing

instructions

received

and a short passage.

second

The

booklet

The memory protocols for

later discarded since their only function was to complete

the

they

regardless of the

would

like

to

see

the

Most subjects

story end in an unexpected way

condition they were in (overall mean = 7.18, SD = 2.44).

This

incidentally, inconsistent with the claim that unexpectedness per se

is aversive (see, e.g.,

they

the

variance

on

expectation

manipulation

expectation

scores showed

was
a

significant effect for level of expectation; F(1, 56)

main

effect

for

valence

approached

significance,

negative versions tended to be rated as less

There

was

expected

= 21.52,
F(1,
than

<

.01.

The

56) = 3.72, p = .06;

positive

versions.

no significant interaction between valence and expectation, (F < 1).

Dunn's multiple comparison tests resulted in significant differences between the

two

positive

and

<
(

(expected vs unexpected) and the two negative conditions, p < .01

.05 respectively.

The mean expectation ratings for the

endings were not significantly different,

two

unexpected

the difference in ratings being a mere

This

finding

is

particularly

important

equally

in the present context

concerning

the

relation

between

affect and expectation require that the level of expectation be the same for the

Responses to the two preending questions were quite uniform.

that

of

because the critical hypotheses to be tested

Preending Ratings

finding is,

A two-way analysis

Evidently,

Insert Table 2 About Here

unexpected.
Results

that

2.

0.73, suggesting that subjects regarded the unexpected endings as being

"cover story."

indicated

Table

for

second passage,

of

of

The instructions told the subject that the experiment

passages, that they would be asked to recall only the

purpose

column

the

passage,

was concerned with memory for what people like to read as opposed to memory
what

The mean expectation ratings are presented in

Unexpectedness ratings.

unexpected/unpleasant respectively.

Each subject received a booklet containing the instructions,

Valence Potential

wanted

Aronson, 1968).

Similarly, subjects strongly

disagreed

to stop before reading the ending (overall mean = 1.34), and

two unexpected endings.

different

either,

The two

expected

conditions

were

not

significantly

although the magnitude of the difference was somewhat

larger

(1.74).

Pleasantness ratings.
Table

2.

The mean pleasantness

ratings

are

also

The two expected endings were both rated as neutral (means:

5.13 for the positive

and

negative

versions,

respectively),

shown

in

4.93 and

supporting

the

none of them chose to do so.

contention

that at

lower levels of the quantitative dimension (unexpectedness),

21

valenced objects can appear

endings

(which

represent

Valence Potential

to be neutral.
high

significantly different (means:

levels

The ratings for the two

of

the

quantitative

unexpected

22

procedure, p <

.05.

Figure 2

shows

Valence Potential

that, as

predicted, the

same

degree

of

dimension) were

7.60 and 3.53 for the positive and the negative

Insert Figure 2 About Here

versions,

respectively).

Thus,

hypothesis, hedonic tone

(i.e.,

contrary

to the predictions of the inverted-U

pleasantness)

increased

with

increments

unexpectedness for positive stories, but decreased for negative ones.

in

A two-way

unexpectedness
story,

high

unexpectedness

analysis of variance on pleasantness ratings resulted in a valence x expectation

story it

interaction

unexpectedness

Figure

1.

F(1,

56)

23.89,$

<

.01.

This interaction is

illustrated in

Consistent with the valence potential hypothesis, the same degree

of

resulted in different degrees of preference.

resulted

in

relatively

influences

low

results

preference,

supporting

the

preference through valence potential.

consistent

Figure 3, subjects in the

the

with

this

positive

positive

in high preference, but for the negative

preending with postending preference data for the two

produced

Insert Figure 1 About Here

resulted

For

unexpected

condition

As

that

Comparison of

unexpected

interpretation.

view

endings

also

illustrated

expressed

in

strong

preference (mean = 8.13) for an unexpected ending prior to actually reading such
unexpectedness

resulted in

valence potential.
valence; F(1,

56)

expectation;

F(1,

or

unpleasantness

depending

on

the

As anticipated, there was also a significant main effect for

= 19.62, p <
56)

two positive (expected

unexpected

pleasantness

versions

.01.

There was no significant

= 1.49; p = n.s.
versus

all

effect

the

significance

two
at

negative,
.01,

and

.05, and

the

two

.01 levels,

respectively.

of

an ending and their level

(mean = 8.07)

unexpectedness

In order to
would

examine

result

in

the

hypothesis

different

that

the

same

degrees of preference,

after

condition, while also

unexpected

Preference Ratings.
degree

Insert Figure 3 About Here

for

Multiple comparison tests pairing the

unexpected),

reached

main

Another

hypothesis

independently

some support from the

postending preference scores.

The main effect for valence, and

the valence x expectation interaction barely missed

3.77
the

for

both,

means

for

p =
the

.06.
two

significance,

F(1,

56)

However, the crucial difference, namely that between

unexpected

endings,

was

significant

using

Dunn's

expressing

for

such

However, subjects
a

high

an

ending

remained

high

in the negative unexpected

preference

(mean

7.20)

for

an

reading one, showed a significant drop in preference

ratings for such an ending after actually reading one (mean = 5.33).

on

direct

preference

reading one.

ending prior to

depending on the valence potential, a 2 x 2 analysis of variance was carried out

the

of

of

its

was

that

effects
fact that

unexpectedness

through hedonic tone.

the partial correlation

influences

This hypothesis received

between

and preference scores, controlling for pleasantness scores

was significant, r = .30, p <

.05.

preference

unexpectedness

(i.e., hedonic tone),

23

at

affect

all,

but

of

sort

preference

and

rules"

been

therefore,

The

preferable."

based

on

as "unexpected things are

such

postending

indirect

The direct postending

included to address this issue.

were

scales

no

experienced

experiment

to the rating scales on the basis of abstract,

"inference

affect-free

interesting,

usually

this

For example, their ratings might merely have

"cold" judgments.
some

in

subjects

responded

scale asked subjects about their preference for a passage with an ending of
same

type as

attempted

measure

to

subjects' preference.
subjects

rate

to

abstracto.

Therefore,

preference

higher

Thus,

story.

this

the influence of direct (concrete, raw) affect on

the

On the other hand,

their

the

reading

for

four

indirect

preference

reflect

ratings

indirect postending scales might serve to increase

scales

required

types of story endings in

different

So these scales were more likely to

judgments.

the

the one they had just experienced, at a time when, presumably, the

affective state they were in was a result of

scale

Valence Potential

preference scale measured subjects preference for a passage with

Direct Versus Indirect Preference Scales

It is always possible that

24

Valence Potential

abstract,

the direct than on the

on
one's

direct scales were indeed measuring experienced affect.

affect-free

confidence

that

pleasant

ending.

Therefore,

for

an

unexpected

this condition, ratings on the direct scale

(mean:

8.07)

are comparable only to

those

on

6.67),

that is, to the scale that asked subjects to rate their preference for a

passage with an unexpected pleasant ending.

(direct

versus

indirect)

the

indirect

Scale

The corresponding

are underlined in Table 3.

pairs

(mean:

of

means

The overall mean for the

direct preference scale (6.32) was significantly higher than the overall mean of
the four comparable (i.e., underlined) indirect scales (5.02),
.05.

The overall mean for the direct preference scale is also comparable

combination

of the overall means of the four indirect scales (i.e.,

the values in the "total" row of Table 3).

also

t(59) = 2.58, p <

significant

at

.05

level.

preference scores was larger than

For
that

This contrast
both

of

tests,

indirect

(6.32

the

vs

5.05)

was

magnitude of direct

scores,

suggesting

that

affective ratings were more pronounced when they were not based on what may have
been relatively abstract ("cold") judgments.

the

Discussion

It should be noted that

conservative assumption that ratings on the indirect

the

scales
two

were not

the mean of

The critical finding with respect to the inverted-U hypothesis is that

this depends on the

to

unexpected

endings, while not rated as significantly different in terms of

influenced by affective reactions to the passage.

unexpectedness, produced

hedonic

tone

ratings

that

were

not

qualitatively

identical, as predicted by the optimal-level theory, but that were diametrically

Insert Table 3 About Here

opposed

to one another.

conclusion
Table 3 shows the mean ratings
indirect

scales,

comparable to

one

corresponding

for
to

these
each

scales.

While

there

were

condition, only one of them was

the direct preference scale that a subject encountered in

his

information

condition.

of

the

critical

intensified affect in terms of the valence potential of

this information.

Rather than directly

itself in terms of

quality.

causing

quality,

quantity

manifested

or
As far as

her

Thus, it appears that unexpectedness

the

preference

results

are

concerned,

subjects

in

the

two

For instance, in the unexpected positive condition, the direct

expected conditions did not differ in their preference ratings.

However, again,

25

those reading the unexpected

of

terms

significant

tone.

hedonic
between

relationship

preference

their

ratings.

but

and

preference

There

ratings.

was

correlation between unexpectedness and preference ratings after the

tone

had

been

partialed

out.

have also been reported by Moynihan (1980), for instance, who

findings

found evidence that pleasure and preference are indeed different dimensions.
might

There are a number of ways in which one

results

in

terms

of

theory.

optimal-level

attempt

to

explain

be argued that other

It might

and

the negative stories interacted with unexpectedness causing the stories in which
the dog was in the back of Marilyn's car (the story for the positive conditions)
generate

an

inverted-U

stories in which
conditions).

Marilyn

was

Optimal-level

curve
a

quite
drug

theory

interaction illustrated in Figure 1 by

Figure

4.

different from that generated by the

dealer

might

thus

the

for
to

two

the

accommodate
curves

for
two

the

correspond

two unexpected versions.

identical

degrees

of

to

the

the

shown

in

degree

of

In this fashion, it might be

unexpectedness

could

is

result

no

in

priori

when

to

the

curve

corresponding

for

while logically possible,

more

are

stories

4,

especially

(curve N) has to be
even

if

the

is

readily

in

degree

In general,

psychologically
explained

the

of

two

this line of argument,

uninterpretable;

terms

pleasantness

we

think

the

of qualitative differences

valence potential rather than in terms of other qualitative differences

in

between

the stories.

Instead

of trying to accommodate the results in terms of other

differences,

an

optimal-level

negative versions differed

factors.

For

instance,

theorist

with
while

respect
the

positive ending, it was perhaps more

counter

such

an

argument.

differ in content complexity.

potential but because of other differences between the two

7 =

complex)

cause different patterns of interaction between unexpectedness and affect.

Figure

Secondly,

obtain

the optimal-level in curve N.

adult judges

would

negative

to

negative stories could be manipulated

results

there

did result in different curves, one would still have to explain how the

ten

that

the

associated with lower degrees of unexpectedness.

required

First,

why the positive and negative stories should generate

as those illustrated in

qualitatively opposite hedonic states, not because of any differences in valence

stories

reason

curves as dramatically different

negative

Point E in Figure 4 might correspond to the degree of unexpectedness

unexpectedness
the

story

attempt

postulating

for the two expected versions and point U would

argued,

(the

There are, however, problems with this kind of explanation.

stories

these

qualitative differences (apart from valence potential) between the positive

to

Insert Figure 4 About Here

However, hedonic tone alone failed to explain the

unexpectedness

Valence Potential

26

were rated as preferable to unexpected unpleasant

contribution of unexpectedness through hedonic

Similar

in

diverged

Thus, unexpectedness seems to influence preference, not directly,

endings.
in

endings

pleasant

Unexpected

endings

Valence Potential

respectively.

might
to

other

negative
complex.

First,

argue

that

qualitative

the

positive and

uncontrolled

quantitative

ending was as unexpected as the

Two

types

of

evidence

would

the positive and negative stories did not

Complexity ratings of the synopses collected from

revealed that the mean ratings on a 7-point scale (1 = simple,

were

4.6

and

4.2

for

the

positive

and

negative

These values were not significantly different.

stories were to be different with respect

to

other

collative

versions,

Secondly, if the
variables,

one

27

such

expect

would

differences

the expected and

judge

tone

basis

in

unexpected

the

hedonic

reasonable to assume that the differences between

preference ratings, it is

stories

were

due

only

to

degree

the

of

unexpectedness

uncertainty

stems

the

and,

is

plausible,

very

that

its influence through tension or arousal reduction.

exerted

is possible that

not

although

possibility,

another

for

the

more

therefore,

two

unexpected

tension.

This

conditions
tension

surprise

reduction

can

convincingly
from tension.

result

in

affect.

positive

Zillmann

tension-reduction

as

the

hypothesis

the positive unexpected

present

results

are

it

could

not

explain

argued

if

the

increment

in

level
only

theorist would be to restrict the scope of

"neutral stimuli"

stimuli

would

for

the

In this case,

constitute a separate source of affective quality.

the past, optimal level theorists have not opted for this solution in
of

embarrassing

or quinine (see,
and

valenced

optimal-

applicability of the theory to

(see Russell & Mehrabian, 1978).

valenced

However, in
the

face

data from studies using clearly valenced stimuli such as sugar

e.g., Walker, 1981).

is

may seem neutral at

Secondly, the distinction between neutral

not always clear-cut.

lower levels of

As we argued earlier, valenced stimuli

the quantitative dimension.

It

be tantamount to abandoning the theory as a general account of the origin

of affect.

Many psychologists, including optimal-level researchers, realize that there

been

problems

no

with

the

alternative

inverted-U

as

broad

If one were

to

hypothesis,

in

scope.

but

The

domains.

It

is

our

contention

unequivocal demonstration of
origin of affect.

that

consistent

potential

the

with

present

be

tied

findings

this

However, although

hypothesis

against

and

were

the

specific

constitute

an
the

of

findings

predicted by it, they do not

The present experiment tested

the

valence

the inverted-U hypothesis, but not against other

in

terms

of

arousal theory, although we find this alternative somewhat

less plausible and more restricted in scope than the

of

to

present

Thus, for example, the results might be explicable

attribution

attribution

to

that there are

We also believe that the valence potential hypothesis carries

hypothesis

alternatives.

continue to use it

the failure of optimal-level theory to explain

the promise of a genuine alternative.

are

the

they

alternatives

(including the attribution of arousal approach) tend

provide unequivocal support for it.

unpleasantness for the negative story.

It seems then, that the only remaining course of action

much.

too

presumably because it has never been definitively refuted, and because there has

even

the

concede

been

were to explain the increment in pleasantness for

condition,

would

have

that positive affect cannot be explained merely in terms of relief

As far

Finally, and perhaps most seriously, from the

are

has

concerned,

our

more

might

(1980)

of the stimuli used in the present experiment merely on the

caused

Optimal-level theorists claim that tension/arousal

was resolved.

valence

perspective of optimal-level theory such a move

It

temporarily heightened by the unexpectedness of the ending and then reduced when
the

the

Valence Potential

of pleasantness and preference scores for the two expected conditions,

stories would seem quite neutral.

would

unexpectedness.

Yet

28

to be reflected in the affective ratings of the

Since expected stories did not differ

two expected versions.

and

Valence Potential

arousal

account

we

present.

account appears to require two assumptions.

An

First, it

must be assumed that unexpectedness exerts its quantitative influence by causing

excitatory

reactions

that

increment

assumed that, lacking an explanation for

initiate

an

epistemic

search

autonomic
their

arousal.

experienced

for its cause (cf.,

Second,
arousal,

Zillmann, 1978).

it must be
subjects
The first

29

assumption seems plausible;

unexpectedness

Valence Potential

but there is some evidence (Iran-Nejad, 1983)

that

can exert its quantitative influence on the experience of affect

by intensifying intellectual activity in the absence of autonomic arousal.

study

In a

similar to the present one, cognitive factors such as interestingness did

not correlate at all with perceived

autonomic

arousal

(r

.07,

30

ns.),

but

Valence Potential

Another way of explaining the data would be to

ratings

in

this

characteristics.
because

it

experiment

were

For example, perhaps

contained

information

merely

the

subjects

that

argue
result

enjoyed

favorably

threatening problems facing a liked protagonist.

that

the

of

the

affective

story-specific
positive

resolved potentially life-

This

interpretation,

correlated highly with surprisingness (r = .43, p < .001) and, most importantly,

conceivable,

with the intensity of affect (r = .40, p < .001).

time subjects encounter the critical conclusion information, Marilyn is

assume

We

are

thus

reluctant

to

that subjects' affective ratings in the present experiment resulted from

cognitive evaluation of unexplained arousal increments caused by unexpectedness.

The second assumption seems untenable as far as the present
concerned.

First,

even

if

subjects

did

experience

experiment

unexpectedness-induced

arousal, it is difficult to see why, having initiated a search for the cause
their

arousal,

of

they would not opt for the most obvious explanation, namely, "I

am surprised and that is why I am aroused."

If they did, they would

to rate the two unexpected endings similarly.

be

likely

But it might be possible to argue

that subjects initiated some sort of a deeper search for clues, in terms of
critical

is

conclusion

information

or the story as a whole.

for instance, that subjects reading the negative

unexpected

the

It might be argued,
ending

determined

in

the

does

not

seem to apply to our results.

safety of a locked office.

back

of

their

minds,

still not be clear why this should be

explained

story-independent

be

explanation

would

indicated

just, legal, and logical.

not

like

it.

that

Discrepancies

of

this

sort

between

cognitive and affective

judgments, though not rare in everyday experience, are difficult to account

given

only

the

assumption

labeling cognitive operations.

that

the

quality

for

of affect arises from arousal-

the

story-specific

it

is

terms.

the resolution of life-

threatening problems per se that is positive because, for instance, it activates

in

the

reader

(observer, etc.) ideas related to safety and security that have

positive valence potential.

interfere

with

these

explain

So long as

ideas,

individual happens to be.

what happened to Marilyn was judged to be moral,

In spite of this, subjects in the main experiment did

in

that

it

Thus,

there

does

even

not
if

is

nothing

matter

the

who

in
the

story-specific

the

ending

to

safe and secure

proposal

could

why the positive ending was rated as positive (as opposed to merely not

generality.

synopses

already

that Marilyn was in danger early in the story, it would

judgment to label their arousal.

the

As Table 1 shows, by the

possibility that subjects were still under the influence of the thought, in

negative), the story-independent account would

rated

though

But even if one were to accept the unlikely

that what happened to their favorite character was "bad" and used this appraisal
However, data collected from judges who

ending

be

preferable

because

of

its

With respect to the unexpected negative condition, a story-specific account

might

propose

that

the

unexpected

contained information that

solved

protagonist

unfavorable

in

manner

the

negative

ending

perceived
to

her.

was disliked because it

problems

The

expected

negative

the

liked

The ending was unpleasant not

because it contained an arrest but because it involved the

protagonist.

facing

arrest

of

liked

ending would be rated as neutral because

31

had

thus

expected negative ending, while

apparent

only

Again we

the initial negativity was intensified and resulted in unpleasantness.

the

prefer

potential

valence

unanimously

rated

critical

the

the

reading

interpretation, it is not clear why the judges

conclusion

the "just deserts"

given

First,

hypothesis.

on

negative

as

information

synopsis

negative

dichotomous scale even though they knew that Marilyn was a drug pusher and
was

why

she

was

terms of sympathy or empathy with story characters must explain why

dislike

both

individuals.

bad

things

Perhaps

negative

valence

disliked

just

and

others

good

regardless

potential

as

like

"things"

some

that

Any theory attempting to explain story affect in

arrested.

(e.g.,

(e.g.,
of

security,

when

things

whether

the

to

bad
have

victims are liked or

compassion)

love,

can

injustice)

rape,

murder,

people

happen

they

are

positive

regardless of whether the beneficiaries are good or bad.

In general,
needs

valence.

cause

differences

in

affective

Perhaps the valence potential alternative can serve as an initial step

towards such a goal as well as

affect

One still

to explain why certain story-specific characteristics (e.g., the degree to

which a reader empathizes with a protagonist)

that

cannot

be

towards

adequately

explaining
clarified

those
in

terms

subjects'

instances
of

of

story

story-specific

For instance, Iran-Nejad

affective

conclusion

ratings

information

were

clearly

Valence Potential

(1983)
not

found

evidence

particularly

resolved

suggesting

that

influenced when critical

protagonists'

problems.

Affective

ratings were much more affected by the degree to which the conclusion provided a
coherent ending to the story, again suggesting that what
valence potential of

set

of

stimulus

is

important

is

the

the information.

An obvious problem with the present experiment is

materials,

leaving

open

the

that it

employed only one

possibility that the observed

effects, although consistent with the valence potential hypothesis, are actually
artifacts

due to a confounding of valence potential and story content.

already argued in our discussion of Figure 4 that

save the inverted-U hypothesis.

such

an

We have

explanation

cannot

However, the fact remains that the positive and

negative stories did differ in content and that this difference corresponded
the

difference

in valence.

It is,

those reported in Berlyne (1

974

reflected

in

such

c) and Schachter and Singer (1962).

studies

and Singer

Portrait
used
to

of

two

different

these

experiments

no

theory

of

the

ratings except when that content

negative

were

stimuli

seem

to

have

been

influence

of

The tests

not

"raw"

one

considered
Perhaps

content

on

is somehow mediated by other factors

(e.g., collative variables in Berlyne's case).

and

enactments,

because of the use of stimuli having different contents.

this is because there is

affective

emotion-inducing

Schachter

induce euphoria and the other designed to induce anger.

of the theories that were provided by

inappropriate

content,

a man and Rubens' Massacre of the innocents.

completely

as

Berlyne used

black-and-white reproductions of two pictures having totally different

Raeburn's

to

of course, impossible to manipulate valence

without some change in content, a fact that is

designed

only descriptive.

story-specific accounts tend to be

properties.

(of

neutrality

As a result, when the ending became unexpected

negative).

actually

something

Thus, the

threatening and therefore intrinsically negative.

of the expected negative ending was

neutralness

been

The valence potential hypothesis, on the other hand, suggests

are

arrests

actually

have

must

neutral,

negative, because in it a bad story character, Marilyn, receives

not

her just deserts.

that

as

rated

this interpretation, the

to

According

her.

for

sympathy

little

positive,

and

condition inferred early on that Marilyn was a drug dealer

subjects in that

32

Valence Potential

In our experiment, the

more

similar

positive

in content than has

33

traditionally been the case.

the

34

Valence Potential

Nevertheless, it would be comforting to know

findings are replicable with different materials.

References

that

More recent work that we

Arkes, H. R.,

have conducted suggests that this

is

in

fact

the

case.

In

the

course

& Garske, J. P.

1983), data similar to

Two basic stories were used

a somewhat different design using a slightly different procedure.

in

In both cases

the equivalent data revealed the same pattern as those reported here.

The main purpose of

theory

this paper has been to demonstrate

cannot account for the origin of affect.

our findings and other evidence

explanation

of

the

(see,

e.g.,

valence

potential

hypothesis.

This,

As an
while

is

that

distinct

Valence
from

alternative,

we

proposed

the

ackn6wledging that cognition can

those

intellectual

potential

affects

affect-specific

responsible for "cold" cognition.

activity)--but

also

organismic
It is

mechanisms

thus independent

the

auditory

properties

are qualitatively distinct from its visual properties.

we have presented disconfirm the optimal-level account of

and

while

they

cannot

be

readily

theories, they are compatible with the

hypothesis.

arousal

of other qualitative stimulus properties

such as meaning potential in the same way that

stimulus

of

determinable valence potential properties of the

not only of quantitative stimulus properties--activity potential (i.e.,

or

no

of

While the data

the origin of

accommodated

by

attribution

predictions

of

the

valence

affect,

of arousal
potential

problems.

T. M. Newcomb,

In

R. P. Abelson,

M. J. Rosenberg,

P. H. Tannenbaum (Eds.), Theories of cognitive consistency:

Crofts,
Berlyne,

&

sourcebook.

Rand McNally, 1968.

Aesthetics

influence

stimulus.

W. J. McGuire,

and

Berlyne, D. E.

affect

of

Progress

In general, our claim, based on

and that affect can influence cognition, locates the origin

the

E. Aronson,

theory:

Conflict, arousal, and curiosity.

1983),

Monterey,

1977.

Berlyne, D. E.

influence affect
in

Dissonance

Chicago:

origin of affect can be based on either arousal itself, or

cognitive labeling of perceived arousal.

Aronson, E.

Wadsworth,

optimal-level

Reisenzein,

that

Psychological theories of motivation.

of
Calif.:

investigating a related, but different issue (Iran-Nejad,

those in the present experiment were collected.

Valence Potential

and

psychobiology.

New York:
New

McGraw Hill, 1960.

York:

Appleton-Century-

1971.

D.

E.

pneumatology

The

vicissitudes

(or

The

Madsen (Eds.),

of

aplopathematic

hydrography of

hedonism).

Pleasure, reward, preference.

New

and

thelematoscopic

In D. E. Berlyne & K. B.
York:

Academic

Press,

1973.
Berlyne, D. E.

Studies in new experimental aesthetics.

New York:

The new

D.

Wiley, 1974.

(a)
Berlyne, D. E.

experimental

aesthetics.

Studies in new experimental aesthetics.

Berlyne, D. E.

Novelty, complexity, and

New York:

grammars.
IX.

& Lichtenstein, E. H.

arousal

Berlyne

In

New York:

Event schemas, story

D.

(Ed.),

(b)
E.

Berlyne

Wiley, 1974. (c)

schemas,

and

story

In J. D. Long & A. D. Baddeley (Eds.), Attention and performance

Hillsdale, N. J.:

Carrol, E. N.,

E.

Wiley, 1974.

interestingness.

(Ed.), Studies in new experimental aesthetics.

Brewer, W. F.,

In

Erlbaum, 1981.

Zuckerman, M., & Vogel, W. H.

theory

of

sensation

Psychology, 1982, 42, 572-575.

seeking.

A test of

the

optimal

level

of

Journal of Personality and Social

35
Experimentelle untersuchungen uber

Engle, R.
unlust

von

der

reizstarke

psychologie, 1928, 64,

beim

abhangigkeit

die

geschmacksinn.

der

lust

und

Archiv fur die gesamte

of

personality.

Springfield,

Illinois:

I.,

&

Tversky,

dimensions.

A.

of

Representations
of

Journal

Experimental

qualitative

Psychology:

and

quantitative
Perception and

Human

&

O'Donnell,

W.

J.

of Experimental Psychology, 1958, 56,

Helson,

Adaptation-level

psychological data.
Iran-Nejad, A.
159).

as

optimal

The schema:

level

a source of affect.

of

Journal

New York:

frame

of

Wiley,

reference

for

prediction
60,

of

Tierney,

P.

Understanding reader's understanding.

(Tech.

Anders,

&

Rep.

J.

N.

Hillsdale, N. J.:

Qualitative and quantitative causes of the experience of affect.

Unpulished doctoral dissertation, University of

231).

Urbana:

February 1982.

Cognition:

University

view.

(Tech

model

for

infant

visual

New York:

Academic Press,

Atkinson,

achievement motive.

Cambridge, Mass.:

From

1975.

Why anyone

would

Poetics, 1980, 9, 87-98.

J.

W.,

New York:
J.

A.

Clark,

R.

W.,

&

Lowell,

Appleton-Century-Crofts,
An

approach

to

E.

L.

The

1953.

environmental

psychology.

M.I.T. Press, 1974.

Structuring process and aesthetic preference in the experience

narrative

forms.

Unpublished

In H. I. Day (Ed.),

doctoral

dissertation,

University

of
of

1980.

The psychological aesthetics of

narrative

forms.

Intrinsic motivation and aesthetics.

New York:

Plenum

Pfaffman, C.

The pleasure of sensation.

Rep.

No.

of Illinois, Center for the Study of Reading,

Psychological Review, 1960,

67,

253-

268.
Pfaffman, C.

Taste

preference

Reinforcement and behavior.

R.

The

Schachter

and

of

the

emotion-eliciting

reinforcement.

New York:
theory

Psychological Bulletin, 1983, 94,

Russell, J. A., & Mehrabian, A.

Illinois, 1983.

A functional

activity

Learning from text, levels of comprehension, or:

McClelland, D. C.,

Reisenzein,

& Ortony, A.

neural

Press, 1981.

1-29.

University of Illinois, Center for the Study

Also in R. J.

Valence Potential

In L. B. Cohen & P. Salapatek (Eds.), Infant perception:

Moynihan, C. & Mehrabian, A.

1949.

Erlbaum, in press.

Iran-Nejad, A.,

California at Los Angeles,

A structural or a functional pattern.

Urbana, Illinois:

(Eds.),

Iran-Nejad, A.

Kintsch, W.

Moynihan, C.

American Journal of Psychology, 1947,

of Reading, February 1980.

Mitchell

the

370-375.

The organization of Behavior.

H.

No.

and

Discrepancy from adaptation level as

Haber, R. N.

Hebb, D. 0.

Infancy

Child Development, 1972, 43, 639-645.

stimulation.

attention.

Mehrabian, A., & Russell,

Performance, 1982, 8, 325-340.

Greenberg, D. J.,

& Maisel, E. B.

read a story anyway.

Thomas, 1967.
Gati,

Karmel, B. Z.,

sensation to cognition (Vol. 1).

1-36.

The biological basis

Eysenck, H. J.

36

Valence Potential

of

In

J.

T.

Tapp

(Ed.),

Academic Press, 1969.

emotion:

Two

decades

later.

239-264.

Approach-avoidance and affiliation as functions

quality

of

an

environment.

Environment and

Behavior, 1978, 10, 355-367.

Schachter, S., & Singer, J.
of emotional state.

Cognitive, social, and

physiological

Psychological Review, 1962, 65, 379-399.

determinants

37

On the psychological law.

Stevens, S. S.

Valence Potential

Psychological Review, 1957, 64,

153-

181.
Thurmond, P. J.
1978,

If

cornered, scream.

Ellery Queen's Mystery Magazine, January,

66-68.

intrinsic motivation and aesthetics.

White,

G.

L.,

Fischbein,

misattribution
1981,
Wundt, W.

41,

In H. I. Day

The quest for the inverted-U.

Walker, E. L.

of

S.,

arousal.

New York:

Rutstein,

&

J.

(Ed.),

Plenum Press,
Passionate

Advances

in

1981.

love

and

the

Journal of Personality and Social Psychology,

56-62.

Grundzuge

der

physiologischen

psychologie.

Leipzig:

Englemann,

1874.
Zajonc, R. B.

Psychologist, 1980, 35,

Zillmann, D.
Harvey,

W.

J.

Ickes,

In J.

H.

& R. F.'Kidd (Eds.), New directions in attribution

Hillsdale, N. J.:

Anatomy of suspense.

functions of television.

American

151-175.

Attribution and misattribution of excitatory reactions.

research (Vol. 2).

Zillmann, D.

Preferences need no inferences.

Feeling and thinking:

Erlbaum, 1978.

In P. H. Tannenbaum (Ed.), The entertainment

Hillsdale: N. J.:

Erlbaum, 1980.

Table 1
Expectation and Valence Potential Manipulations
as a Function of the Critical Stem and Ending Informationa

Critical Stem Information

Condition

Critical Ending Information

Negative
Unexpected

Expected

A police car is behind Marilyn.

Marilyn is perhaps going to be raped/mugged.
Gabriel is probably a wolf in sheep's clothing.

A police car is behind Marilyn.

Marilyn is perhaps going to be raped/mugged.
Gabriel is probably a wolf in sheep's clothing.
Marilyn is
Gabriel is
Marilyn is

Marilyn is not raped/mugged.

Gabriel is not a wolf in sheep's clothing.
Police discover heroin in Marilyn's car.
Marilyn is a drug dealer.
Gabriel is a police informer.
Marilyn, a guilty person, is arrested.

perhaps a drug dealer.

probably a police informer.
perhaps going to be arrested.

Positive
Unexpected

Expected

Marilyn is probably going to be raped/mugged.

A rapist/mugger is in the area and could be in
Marilyn's car.
Gabriel is probably a Good Samaritan.

Marilyn is probably going to be raped/mugged.

A rapist/mugger is in the area and could be in
Marilyn's car.
Gabriel is probably a Good Samaritan.

Marilyn is not raped/mugged.

No rapist/mugger is in the car.
Attendant's friendly dog emerges from the car.
Marilyn was never in danger.
Gabriel thought a rapist/mugger was in the car.
Gabriel acted to save Marilyn.
Cabriel is a Good Samaritan.
Marilyn, a good person, is safe and secure.

Hospital attendant has lost his (friendly) dog.

Dog often sleeps in attendant's car.
Attendant's dog may be hiding in Marilyn's car.
I - -- ---I-~-~-"~---"~-~-"--"-- ~- -- '
al'he information in this table does not necessarily rep)resent information explicitly stated in the text.
it represents the gist of the most relevant, i.e., critical, information.

---~-Rather ,

Table 3
Table 2

Mean Ratings for Direct and Indirect

Mean Expectation and Pleasantness Ratings by Condition

Expectation
Condition

Ratings

Pleasantness
Ratings

Postending Preference Scales

Preference Ratings
Condition

Direct

Indirect

Indirect

Scale

Scale 1

Scale 2

Indirect
Scale 3

Indirect
Scale 4

Mean

S.D.

Mean

S.D.

Expected Positive

6.07

3.56

4.93

2.05

Expected Positive

4.93

7.13

4.33

4.07

4.67

Expected Negative

4.33

2.87

5.13

1.55

Expected Negative

5.93

5.80

4.67

4.00

6.07

Unexpected Positive

2.60

1.45

7.60

1.81

Unexpected Positive

8.07

6.67

3.80

4.00

4.07

Unexpected Negative

1.87

1.25

3.53

1.25

Unexpected Negative

5.33

6.93

5.40

4.07

5.07

6.32

6.63

4.55

4.03

4.97

Total

Figure Captions

Figure 1.

Mean hedonic tone ratings for positive and negative story

endings as a function of level of expectation.

Figure 2.

9 8-

Mean preference ratings for positive and negative story

7-

endings as a function of level of expectation.

Figure 3.

Mean preference ratings for the two unexpected story endings

as a function of time of judgment.

Figure 4.

Hypothetical relationship between unexpectedness and hedonic

tone resulting in separate curves for the negative (Curve N) and positive
(curve P) story versions.

0
IQ

LU

6E

5 - 5432-

I
0

EXECE

EXPECTED

----

POSITIVE
NEGATIVE

UNXECE
UNEXPECTED

9.

9.

8-

7.

W 6
z
a: 5

w 6z
w
Lcr
w 5
LL
W
a: 4-

()

LL
1:
a.

4-

3-

3 -

2-

POSITIVE

I -

--

O -

i
EXPECTED

UNEXPECTED

POSITIVE
NEGATIVE

NEGATIVE

O-

PREENDING

POSTENDING

POSITIVE
HEDONIC
TONE

INDIFFERENCE

NEGATIVE
HEDONIC
TONE