RD-AI68 26?

BETWEEN HOPE AND FEAR: THE PSYCHOLOGY OF RISKCU) 1/1

NISCONSIN HUMAN INFORMATION PROCESSING PROGRAM MADISON
L L LOPES MAY 86 NHIPP-24 N96914-84-K-SS65

UNCLASSIFIED F/G 5/1 M

IIEEEEEEEEEEEE
IIflflfllflflflf
EEEEEEEEE
 mini
LI M

1 2 4 01.

1160''*~ -

-i - -" -

' ll2 " *'

MICOcopy RESOLUTIN TES4CHAR1

NA .I

p I 'II'

p ~. ~. .
pt
CHA.
MICRCOPYRESOUTIN TFSI

. .. . . . .. . . . . . . . . . . . . . ... . ,Ji', t, .

-Lo

" "
-- .- -- .

i*.j
... 44
 BETWEEN HOPE AND FEAR:
0 THE PSYCHOLOGY OF RISK
00
LOLA L. LOPES

WHIPP 24 MAY 1986

This research was supported by the Engineering
Psychology Program, Office of Naval Research,
ONR Contract N0001 4-84-K-0065, Work Unit NR 197-079

Approved for public release; distribution unlimited.

Reproduction in whole or part is permitted for ~
any purpose of the United States government. "W
A7
WISCONSIN HUMAN INFORMATION PROCESSING PROGRAM
DEPARTMlENT OF PSYCHOLOGY uUNIVERSITY OF W4ISCONSINI MAlDISON, WISCONSIN 53708

~'l FILE COI?

.U6 -'
 unclasslrled
SECURITY CLASSIFICATION OF THIS PAGE (lWhen Date Entered) _

READ INSTRUCTIONS
REPORT DOCMENTATION PAGE BEFORE COMPLETING FORM
. REPORT NUMBER 2. VT ACC No N
. CATALOG NUMBER

WHIFF 24 .y40!s
4. TITLE (ad Subtitle) S. TYPE OF REPORT G PERIOD COVERED

Between Hope and Fear: Interim Report

The Psychology of Risk 6. PERFORMING ORG. REPORT NUMBER 4%

7. AUTHOR(a) li. CONTRACT OR GRANT NuMiuERe)

Lola L. Lopes NO0014-84-K-0065

3. PERFORMING ORGANIZATION NAME AND ADDRESS 10. PROGRAM ELEMENT. PROJECT. TASK
A W N U
The Board of Regents of the
University of Wisconsin System NR 197-079
750 University Ave., Madison, WI 53706 N17-9
I. CONTROLLING OFFICE NAME AND ADDRESS 12. REPORT DATE

Office of Naval Research May 1986_

800 North Quincy Street 13. NUMBER OF PAGES

Arlington, VA 22217 61
14. MONITORING AGENCY NAME • AODRESS(If different from Controlling Office) IS. SECURITY CLASS. (of thle report)

Unclassified
ISa. DECLASSIFICATION/DOWNGRADING
SCHEDULE ,,-.**

IC. DISTRIBUTION STATE CNT (of thl Report) * ". *.

Approved for public release; distribution unlimited.

17. DISTRIBUTION STATEMENT (of the abstirac ntered in Block 20. It different tram Report)

IS. SUPPLEMENTARY NOTES .:. -

IS. KEY WORDS (Continue on reverse sde If necessar and Identify by block number)

risk.
decision making_,
motivation,

20. ABSTRACT (Continue on reverse side If neceesary and Identify by block number)

Psychophysical theories of risk (e.g., Bernoullian utility theory,

prospect theory) are compared with motivational theories (e.g., those
of McClelland and Atkinson). A new theory is proposed that describes
risk taking in terms of two factors: Factor 1 is a dispositional
variable involving motivation for security versus potential. This
factor distinguishes risk averse and risk seeking individuals and

DO I , 1473 EDITION OF I NOV I OSOLETEUnclassified

5 N 0102- LF- 014-6601 Y CLASSIFICATION
SECURITY e
classifiT OI" THIS PAGE (Whe Data Entered) ,T: _

%, .''
%-7 " PI"
6 _Unclassified
Wcumyrv cLAmnCATMW oF T pAeC CMW D", -ih4

reflects primarily whether the individual is motivated to avoid bad

outcomes or to achieve good outcomes. Factor 2 is a situational variable
involving current level of aspiration. Level of aspiration reflects
immediate needs and opportunities. In the theory, these two factors
are sometimes in correspondence and sometimes in conflict, predicting
complex patterns of data. Evidence is presented to support the theory
and the relationship of the theory to other concepts (e.g., safety-first,
disappointment, regret, emotion, aesthetics) is discussed. -> -.

44

Sr I.

.1.

., ** **.* '* p -,o

-p.. I

Si/M0102- IF-Ol14- 6601

Unclassified
SECUSIYY CLASMflCAflWO OF ThIS PA6Et~b 3.f Xnrm4
 Between Hope and Fear:
The Psychology of Risk

Lola L. Lopes
Department of Psychology
University of Wisconsin
Madison, Wisconsin

"-....

In press:
Advances in Experimental Social PsycholQyv

Aceeson Fo.
NT I.

D)TIC TA~
. . -- - "
Justl'["'

Dislri '

,°- .

OIALITY
NSPECTILOj
30
a ~% .****. ~ ' -U
 Psychology of Risk / 2

(Contents)
I. Introduction
A. The Experimentalist's View
B. The Personologist's View
II. Psychophysical Versus Motivational Theories of Risk
A. The Role of Cognition
B. Risk-Taking Reconsidered

III. The Task and the Representation of the Stimulus

A. How Can We Represent Risks?
B. How to Draw a Lorenz Curve (and Why)
IV. A Two-Factor Theory for Risky Choice
A. Factor 1: Security Versus Potential
B. Factor 2: Aspiration Level
C. Conflict Between Security/Potential and Aspiration
V. Evidence for the Two-Factor Theory
A. Riskiness Is the Absence of Security
B. Risk Attitude Is More Than the Psychophysics of Money
C. Risky Choice Is Not Conflict Free
VI. The Things We Don' t Talk About
A. Fear and the Safety-First Principle
B. Planning Is Applied Hoping
C. Anticipation and Imagination
D. Sherpas and Other High Rollers

OleI

".°.

,'"""'- - '..
","%
.'.v"".."..
.v .' . . . ."-".. . .,, "v, -" ,-".
. -'.v.v. .. ..,... ..,." ,. ."-
.,,&,[. -, ..,..,,, , ", "U'-'
_ _,'...x,"
 Psychology of Risk / 3

Most things begin to look a little funny if you stare at them long
enough. So too the psychology of risk. What is most disconcerting is that
there is so much theory for so little substance. Countless hours have been
spent by psychologists and economists alike in trying to explain theoretically
why people buy both lottery tickets and insurance. Lottery tickets cost a
dollar. One. We buy insurance (when we can afford it) so that we can sleep
better. Is it really so strange that we should want to buy both?

This article is about risk: what risk is (if it is any thing at all);
how people think about it; what they feel about it; and what they do about it.
The article is also about how psycholoqists think about risk: how they study
it; what tasks they use; what factors they vary; and what models they build
(or borrow) to describe people's behavior.

Technically, the word risk refers to situations in which a decision is

made whose consequences depend on the outcomes of future events having known
probabilities. Choices among the different kinds of bets in games like
roulette and craps are good examples of choices made under risk. Insurance
companies also operate under risk when they set the premiums for ordinary life
insurance. But most of the time our knowledge of probabilities is not so
exact. Sometimes it's pretty good (as with the weather tomorrow or the going
rates for auto loans just now); other times it's pretty awful (as with whether
a wedding reception should be held indoors or outdoors several months hence or
whether the fixed rate mortgage that's offered today is going to feel like a
bargain or a burden 10 years down the road). When our knowledge of
probabilities is very inexact (or lacking entirely) we say that decisions are
made under uncertainty or ignorance. Obviously, risk shades into ignorance
and most important decisions are made part way between the poles.

Psychological studies of risky choice (which 4,3 the term used

conventionally to refer to all but the most extrere instances of ignorance or
ambiguity) fall into two groups. At one extreme are the studies run by
mathematically inclined experimental psychologists. These are studies in
which subjects make decisions about gambles which are described in terms of
amounts and probabilities. At the other extreme are studies run by
personality psychologists who are mostly interested in individual differences
in risk taking. Their tasks tend to be closer to everyday experience and they
often involve elements both of chance and skill.

A. 7he 'Alists' View

A good example of an experimental task comes from Kahneman and Tversky

(1979) who asked subjects questions similar to this: which would you rather
have, $3,000 for sure or an 80% chance ol winning $4,000? Mcst subjects
prefer the $3,000 for sure even though the expected value of the gamble is
V. higher, .80 x $4,000 = $3,200. Such preferences are conventionally labeled
a 45% averse"
"risk chance ofaswinning $6,000. favoring a 90% chance of winning $3,000 over
are preferences

w Experimental psychologists tend to explain risk averse behavior in one of ,

two ways. Some theories of risky choice see the subject as trading off
two ptr

L.
 Psychology of Risk / 4

potential return with "risk," a construct that is most often identified with
variability in the outcome distribution. For example, Coombs's (1975)
"portfolio theory" is based on the premise that choices among risks reflect a
compromise between maximizing expected value and achieving an individually
determined ideal level of risk. Thus, investors who prefer low risk must
accept the lower but safer returns associated with bonds whereas investors who
prefer more risk can opt for the higher but less safe returns of stocks and
commodities.

Theories that consider risk to be a function of the variability among

potential outcomes are intuitively appealing, but they are less commonly held
than alternative theories based on weighted value models, the best known of
which is expected utility theory. In expected utility theory, subjects are
assumed to "compute" something similar to an expected value, but instead of
using the objective dollar amounts, they operate on subjective amounts (or
utilities) which are usually nonlinearly related to dollar amounts.

The first use of the utility concept was made by Daniel Bernoulli in
1738. He argued that the value of money is not absolute, but depends on how
much one has already: "Any increase in wealth, no matter how insignificant,
will always result in an increase in utility which is inversely proportionate
to the quantity of goods already possessed" (Bernoulli 1967, p. 25). This
view implies that a constant gain, say $1000, will be worth more subjectively
to a poor person than a rich person. It also implies that subjective
differences between amounts of money that differ by a constant get smaller as
the absolute magnitudes of the amounts get larger. In other words, there's
more difference psychologically between $1,000 and $2,000, say, than betweeen
$10,000 and $11,000. In Bernoullian terms, money has marqinally decreasing
utility (which is short for saying that the subjective value of constant
increments gets smaller and smaller). In mathematical terms, the utility
function (which relates the objective value of money to its subjective value)
is negatively accelerated. An example of a negatively accelerated utility
function is shown in the upper left quadrant of Figure 1.

Insert Figure 1 about here

Bernoulli's utility theory predicts risk aversion. This is easy to

understand intuitively. Consider a gamble that offers a .5 chance of winning
$2,000 and a .5 chance of winning nothing. The expected value of the gamble
is $1,000, but it is worth less subjectively to anyone having a Bernoullian
utility function. This is because the subjective value of $1,000 is more than
half the distance between nothing and $2,000. Thus, the utility of $1,000 for
sure must be greater than the averaae of the utility of nothing and the
utility of $2,000 (which is all that the expected utility is -- an average in
which utilities are weighted by their probability of occurrence).
A problem, however, is that people are not always risk averse. For
example, they buy lottery tickets even though it is well known that, on
.verage, lotteries are mathematically unfair (which doesn't mean, of course,
that they are crooked, but only that the price of the ticket is more than the
expected value). People are also sometimes "risk seeking" (as it is
conventional to say) when choosing among risks that have negative consequences
 Psychology of Risk / 5

(Markowitz, 1959; Williams, 1966; Kahneman & Tversky, 1979). For example,
many people would prefer to face an 80% chance of losing $4,000 than to lose
$3,000 for sure even though the expected value of the risky loss ($3,200) is
greater than the sure loss.

Risk seeking choices can be explained by weighted value theories in two

different ways. One way is to say that the utility function is not always
negatively accelerated, but instead has one or more regions of positive
acceleration. Three examples of such kinky utility functions are shown in
Figure 1. The function in the upper right quadrant was proposed by Friedman
and Savage (1948), that in the lower left by Markowitz (1952), and that in the
lower right by Kahneman and Tversky (1979A. Each of these functions has one
or more regions of negative acceleration (which predict risk averse behavior
such as the purchase of insurance and the rejection of gambles) and one or
more regions of positive acceleration (which predict risk seeking behavior
such as the purchase of gambles, notably long shots that offer very large
prizes at very small probabilities).

Another way to explain risk seeking choices is to suppose that the

weights people use in evaluating gambles are not identical to the objective
probabilities. Thus, people may purchase lottery tickets because they
(optimistically) assign more weight to the probability of winning than is
justified objectively. This notion that subjective probabilities may differ
from stated probabilities is consistent with subjectively expected utility
theory (Savage, 1954). It is also consistent with more general weighted
utility models (Karmarkar, 1978; Kahneman & Tversky, 1979) in which the
subjective values attached to probabilities don't even act like probabilities.

To sum up, then, most experimentalists explain risky choice by positing

an internal process for evaluating gambles that is structurally similar to
computing expected values. However, the objective values are replaced by
subjective values such as utilities and probability weights that differ
quantitatively from their objective counterparts.

B. 7Te Prs logists' View

A good example of a task used by personality psychologists is the "ring

toss" game originally used by McClelland (1958) in his studies of achievement
motivation. In this task, subjects throw rings onto a peg from a distance
that they are allowed to choose themselves. At one extreme, they can stand so
near the peg that success is virtually certain. At the other extreme, they
can stand so far away that success is virtually impossible. The variable of
interest, therefore, is not how well they do objectively, but rather where
they stand.

This is a task in risky choice (even though the probabilities are only
roughly known) because subjects are choosing among alternative actions that
have the character of gambles. One can choose (and some do) to stand near the
peg in which case the probability is high, but the personal satisfaction in
getting the ring over the peg is small. Alternatively, one can stand further
back in which case the probability is smaller but the satisfaction is greater.
One can even choose a long shot (literally) in which the probability is
virtually nil, but the potential satisfaction is enormous. (Consider the

:',[,. '- '. - -i? -'' ¢ ...-

.,,-,%V -?.?-',?.
,' ".'-.-.--i .:.<.-:.-:..-- ..". <? ?-i--. -:-::.:..:.-.. ; - ;:- '?-;--'- -.. - .-.
 Psychology of Risk / 6

heady burst of satisfaction that comes in basketball when a desperation shot,

tossed the full length of the court, drops neatly through the net. It helps,
of course, if it is one's own team that has made the shot.)

Personologists explain people's choices in the ring toss game in term of

their motivations (McClelland, 1961; J. Atkinson, 1983). People who have high
motivation to succeed (M ) tend to pick intermediate distances (particularly
if they are also low in the motive to avoid failure, M f). The theoretical
rationale for this is that there is a three-way multipficative relation
between motivation, probability, and incentive (J. Atkinson, 1957). Since
probability of success (closeness to the peg) and incentive value (anticipated
satisfaction from making the particular shot) are inversely related to one
another, the attractiveness of the task for the person with high M and low
M f s least when the probability of succeeding is either very small jr very
19ge, and it is maximum when probability is at an intermediate value

People who are low in Ms, however, especially if they are also high in
M , more often stand either very near or very far from the peg. Either way,
tAse people reduce their performance anxiety by virtually guaranteeing that
they will experience either the tepid success of a trivial accomplishment or
the denatured failure of not doing something that can't be done anyway except
by sheer good luck.

Thus, personologists focus on the similarities and differences between

different people's needs and on the various options that people pursue for
meeting their needs. They also stress the often competitive nature of needs
and the resulting compromises that must 1ce made between certainty and
incentive, for instance, or between the motive to achieve success and the
motive to avoid failure (both of which can exist in the same person at the
same time.)

I. Psychcphysical versus mtivational Theories of Risk

The gamble task and the ring toss task appear to have little in common,
but they are alike structurally. In each case the subject chooses among
alternative options, each of which can be characterized by the probabilities
and values that are attached to uncertain outcomes. In fact, one can with
some justice interpret subjects' choices in the ring toss task as maximizing
expected utility (cf. J. Atkinson, 1958). But there are profound differences
in the theoretical mechanisms that have been used to explain subjects' choices
in these two tasks.

Bernoullian theories are psychophysical in exactly the sense of ordinary

' sensory psychophysics. In fact, the quantitative function that Bernoulli
suggested for utility reappeared more than 100 years later as Fechner's
general psychophysical law. But this means that in Bernoullian theory, risk
* averse behavior has neither to do with risk nor with aversion. Instead, it
has only to do with the way we experience quantity. The person who turns down
an 80% chance of winning $4,000 in favor of a sure $3,000 is not, therefore,
avoiding risk, nor even experiencing risk in any theoretically relevant way,
but is only responding to the same sorts of factors that make the difference
in heaviness between 1 pound and 2 pounds seem greater than the difference

-- ~ * S.,

+, ,. -++ jl.+ll.
, i ***-*_ *-* -- 9 * * ), * * * "
 Psychology of Risk / 7

between 10 pounds and 11 pounds.

The same can be said for theories in which objective probabilities are
replaced by subjective probabilities or by probability weights. In fact,
Kahneman and Tversky (1984) have referred to their decision weight function as
reflecting the "psychophysics of chances" (p. 344). Their decision weights
are assumed to differ from objective probabilities due to factors that are
like those found in perception. For example, category-boundary effects make a
change from impossibility to possibility (or from possibility to certainty)
seem larger than a comparable change in the middle of the scale. This is the
same sort of thing that happens in the "categorical perception" of human
speech. Sounds that actually change in equal acoustic increments from an
ideal "ba" to an ideal "da" are perceived to shift abruptly from sounding like
clear ba's to sounding like clear da's (Liberman, Harris, Hoffman, & Griffith,
1957).

Motivational theories, on the other hand, construe the decision maker as

analyzing possible outcomes and assessing risks (by which is usually meant
simply the probability of achieving some goal). For example, McClelland
(1961) gave the following account of how children who are high in achievement
motivation decide where to stand:

If they stand too close to the peg, they are much more likely to
throw the ring on.. .but they are less likely to get any feelings of
achievement satisfaction from doing so. If they stand too far away,
they are both much less likely to succeed and more likely to regard
success as "luck," than if they stand a moderate distance from the
peg. In fact, they are behaving like the businessman who acts
neither traditionally (no risk) nor like a gambler (extreme risk),
but who chooses to operate in a way in which he is most likely to
get achievement satisfaction (moderate risk, in this case about one
chance in three of succeeding) (p. 212).

There are many places in McClelland's account where one could evoke
psychophysical mechanisms to explain why children make different choices:
they might attach different subjective values to success at different
distances; or they might assign different probabilities to succeeding at a
given distance; they might even differ in the way they perceive distances.
But this is not how achievement theorists explain things. McClelland is
careful to rule out such factors. For example, he acknowledges that children
who are high in achievement motivation tend to perceive their probability of
success more favorably than children with low achievement motivation,
particularly when there is no evidence one way or the other. However, when
they have reasonable knowledge based on past performance, they use that
knowledge appropriately and do not display a greater perceived probability of
success than children who are low in achievement motivation. In his words,
achievement motivated children "are not impractical 'dreamers' overestimating
their success at everything; instead they rely on facts so far as they are
available, and then fall back on generalized self-confidence" (McClelland,
1961, p. 223).

One can also examine the possibility that achievement motivated children
simply place higher subjective value on success by looking at their
preferences in games of pure chance. In this situation they clearly prefer

V -- * . ~ . . . - .. . . . . . . . . . . .
 Psychology of Risk / 8

the shortest odds they can get (the safest bets) followed by intermediate
values. Children with low achievement motivation, on the other hand, tend to
like long shots that offer large prizes, but at small probability (McClelland,
1961).

A. 7 e Role of Cognition
We have, then, two different approaches to explaining risky choice, one
primarily psychophysical and the other primarily motivational. The theories
differ also in the degree to which they are coQnitive. The motivational
theories have strong cognitive components. In order to choose appropriately,
task difficulty must be analyzed for the relative contributions of skill and
chance, past experience must be marshalled and used to assess probabilities,
goals must be set and future feelings predicted about what will be satisfying
and what not. Thus, it is motivation that incites action and gives it
direction (i.e., approach or avoidance), but it is cognition that guides
action to its intended goal.
Psychophysical theories, on the other hand, have not been couched in
cognitive terms, although they certainly might be. One could, for example,
justify a Bernoullian utility function in terms of Maslow's (1954) notion of a
need hierarchy. On this view, $1,000 really is worth more to a poor person
than a rich person because the poor person will spend the money to satisfy
more basic needs (food, shelter) whereas the rich person will spend it on more
transcendent needs (operas, electronic running shoes). Likewise, people's
tendency to treat small probabilities as zero might be justified cognitively
in terms of th, degree to which small probabilities can be expected to produce
discernable impacts on how we choose to live our daily lives. This was done,
in fact, by the early probabilist, Buffon, who urged that all probabilities
less than .0001 be treated as "morally" (which is to say, psychologically)
equal to zero (Daston, 1980).

This is not, however, how psychophysical theories are justified. Indeed,

they tend most often not to be justified at all. But Kahneman and Tversky
(1979) have been refreshingly clear about their theoretical foundations. They
say of their value (or utility) function,

Our perceptual apparatus is attuned to the evaluation of changes or

differences rather than to the evaluation of absolute magnitudes.
When we respond to attributes such as brightness, loudness, or
temperature, the past and present context of experience defines an
adaptation level, or reference point, and stimuli are perceived in
relation to this reference point ....Many sensory and perceptual
dimensions share the property that the psychological response is a
concave function of the magnitude of physical change ....We propose
that this principle applies in particular to the evaluation of
monetary changes (pp. 278-279).

Likewise, their description of probability weighting, while less clearly

articulated, seems to rest primarily on perceptual and attentional metaphors.

7
 -~ -7 -7 - LITlK w-Tx winu TV WV NV-- -7 . 4. -. 7. W -7

Psychology of Risk / 9

B. Risk-Taking Recnsidered

Personality psychologists and experimental psychologists tend to have

very different goals. Personologists, at least traditionally (i.e., pre-
Mischel, 1968) have taken an idiographic approach to explaining behavior.
Thus, they have been concerned with the things that make us different from one
another. Since these are necessarily attached to the individual and not to
the situation (which is held constant), the theoretical emphasis has fallen on
the structures and dynamics of the inner person. Experimentalists, on the
other hand, have typically taken the nomothetic approach which is aimed at
understanding Everyman and the ways in which we are all alike. This approach
puts the emphasis on the comonly experienced environment which, in the
context of the laboratory, reduces to the stimulus. Experimental theories,
whether they are behavioral or cognitive, tend to revolve around the
transformation of the stimulus into the response. If this can be done without
invoking individual differences or higher level cognition, all the better.
Hence the appeal of the psychophysical metaphor for explaining risky choice
(or more properly, for explaining the most common choice pattern while
ignoring entirely the patterns of a substantial minority of subjects).

When the scientific paths of personologists and experimentalists cross as

they have in the area of risky choice, the weaknesses of each tradition are
illuminated by the strengths of the other. Personality theorists paint with a
broad brush and a richly hued palette, at least compared to their
monochromatic experimental colleagues. McClelland's (1961) The Achieving
Society is breathtaking in its scope and intent, ranging methodologically from
the laboratory to the field and substantively from history to economics and
from psychology to sociology. But the experimental evidence tends to be
unsystematic and unconvincing, at least for the experimentalist schooled in
the parametric (if you can vary it, vary it) tradition. Thus, although the
motivational approach is appealing for its whole-person flavor (with
motivation and emotion and cognition all having their place), the actual
experiments are scattered far too sparingly over the conceptual domain. In
particular, the motivational treatment of risk-taking in the domain of pure
chance is disappointing, especially when one considers that we regularly deal
with risks (e.g., farmers planting crops, investors choosing a stock) whose
outcomes are largely out of our personal control.

Experimentalists, on the other hand, tend to explore their domain more

thoroughly, not necessarily because they are better scientists, but because
the tradition of looking for "critical tests" keeps the experimental stimuius
changing in interesting ways. Thus, the history of thought in risky choice
has proceeded in relatively discrete steps as paradoxical results posed
initially as challenges to the theory eventually became accomodated via
theoretical elaboration. (A comparison with Ptolemaic astronomy would no: oe
unwarranted after 250 years of elaboration on a theory that has remained
essentially unchanged structurally). But in the service of the detaileJ -.ew,
the big picture tends to be lost. So it is with risky choice; after all the
study and all the clever theorizing, we are left with a theory of risk-:akin:
that fails to mention risk. It also fails to consider (much less explain) the
motivational and emotional factors that give risky choice its experentiall
texture: the hopes and fears that give us in due measure both purpcse and
pause.
 Psychology of Risk / 10

In the remainder of this article, I present a theory of risky choice that

attempts to meld the strengths of both approaches. Empirically and
methodologically it is tied to the experimental approach to risky choice. But
theoretically it is more strongly tied to motivational approaches,
particularly those of McClelland (1961) and J. Atkinson (1983).
Nevertheless, the theory was developed independently of the latter theories
and has at least some formal roots in economic3 (see Lopes, 1984). Although
the basic theoretical constructs of the new theory are quite similar to those
found in the achievement literature, I will make no particular attempt to
bring the two approaches into tighter theoretical alignment since that could
(and probably would) do disservice to the fact that the task domains have
important differences, particularly those involving the skill/chance
dimension. However, the strong theoretical similarities increase my
confidence in both approaches.

III. The Task and the rs tin of the Stimulus

The term "risky choice" can be read two ways. Risky choices are choices
that have an element of danger. They are risky and may come to a bad end.
Losses may be sustained, hopes may be shattered, or opportunities wasted.
Risky choices are also choices between risks or between risks and sure things.
In this sense, risks are gambles. Most research on risk has concentrated on
gambles in which there are only two possible outcomes. In fact, a not
inconsiderable part of this research has dealt with what might be called one-
outcome gambles in which one outcome represents a change (e.g., winning
$4,000) and the other represents the status quo. The focus on two-outcome
gambles seems reasonable to most researchers in part because such gambles lend
themselves well to parametric manipulation in the laboratory. In addition,
two-outcome gambles are conceptually simple, a fact of at least some
consequence 3 given the known limitations in human beings' ability to process
information
Real-world risks, on the other hand, hardly ever have just two outcomes.
More often they range essentially continuously over the outcome variable.
(Consider interest rates on Individual Retirement Accounts. As I write,
they' re averaging from around 7.5% for short term investments up to around
9.5% for long term investments. Is 9.5% enough to tie up funds for a long
period, or will we have another bout of high interest rates? And if so, how
high will they go?) In fact, two-outcome gambles occur mostly in the context
of formal gambling (and psychology experiments). A $2 bet on red in roulette,
for example, will either win $2 or lose $2. Likewise, a horse player betting
$2 at odds of five to one will either win $8 or lose $2. In either case,
however, it seems unlikely that players will make a separate decision each
time they place a bet. Instead, the decision to play usually entails placing
a series of bets, with resulting net outcomes that range in principle from all
losses to all wins.

S %~
.. ~ ig . . ~ .. . * * *-.. -..- * . . . . . •. •j.. . .-.- o|
 Psychology of Risk / 11

Insert Figure 2 about here

The experiments described in this article have investigated how people

chose among multioutcome gambles (or "lotteries" as we refer to them with
subjects). Figure 2 gives six examples of these lotteries listed in the order
in which they are preferred by risk averse subjects (Schneider & Lopes, in press):
riskless > short shot > peaked > rectangular > bimodal > long shot. Each of
the lotteries has 100 lottery tickets (represented by tally marks) and each
has an expected value of approximately $100. The lotteries differ, however,
in how the prizes are distributed. The long shot, for example, has 31 tickets
that each win nothing, 22 tickets that each win $49, and so forth up to a
single ticket that wins $439. In contrast, the short shot has only 1 ticket
that wins nothing, 2 tickets that each win $13, and so forth up to 31 tickets
that each win $130. (Note that the riskless lottery is so named because it
has a riskless, i.e., sure, component that guarantees, in this case, a minimum
win of $70.)

All the lotteries in the figure are "gain" lotteries which means that
their prizes are all > 0. Loss lotteries were also used in some of the
experiments. Loss lotteries are just like gain lotteries except that their
outcomes are negative. Thus, for example, the long shot for losses has 31
tickets that each lose zero, 22 tickets that each lose $49, and so forth down
to 1 ticket that loses $439. Likewise, the riskless loss lottery guarantees a
riskless (sure) loss of at least $70.

Three kinds of task have been used. In the most common task subjects
were shown pairs of lotteries in all possible combinations and asked which
they would prefer if they were allowed a free draw from either (Lopes, 1984,
Experiments 1 and 2; Schneider & Lopes, in press). Pair-preference data can
be used to infer preference orders over the entire set of stimuli. The second
kind of task involved judgments of riskiness also expressed in pair-choices
(Lopes, 1984, Experiments 3 through 6). These can be used to infer risk
orders (which are not the same as preference orders except for risk averse
people). The third kind of task (Lopes, 1986) was embedded in a standard
pair-preference task. Subjects were shown pairs of lotteries and asked to
express their preferences for each. For a subset, however, they were also
asked to explain their preferences in writing. These written protocols were
collected for a group of 14 graduate students from a variety of departments.
In order to avoid the known pitfalls of retrospective reports (Ericsson &
Simon, 1980), the protocols were obtained directly at the point of choice.

The present article focuses primarily on preference data from Lopes

(1984) and Schneider and Lopes (in press). Examples of the verbal protocols
are used throughout, however, for illustration. In the protocols lotteries
are referenced by the names listed in Figure 1. These names were not,
however, used by the subjects.

'
." " . t . . * . *. . .
..-- . . .. . . . . . ." - f. ..
 Psychology of Risk / 12

A. Ebw Can We PRqesaeit Risks?

One of the most important steps in psychological theorizing is to find a
representation of the stimulus that has psychological fidelity, by which I
mean a representation that highlights the stimulus features that actually
affect behavior. For the most part, two-outcome gambles have been treated as
though the functional stimulus is identical with the presented stimulus: a
pair of outcomes each associated with a probability of occurrence. For
lotteries like those in Figure 1, however, we seem to respond more to the
.shave of lotteries than to the amounts and probabilities of individual
outcomes.

The idea that risk is a function of shape has been proposed both for
theories of risk perception (Luce, 1980; Pollatsek & Tversky, 1970) and for
theories of risk preference (Allais, 1979; Coombs, 1975; Hagen, 1969;
Markowitz, 1959). In these theories, shape is identified with the statistical
moments of the distributions, particularly mean, variance, and skewness
Variance is generally considered to be bad (i.e., risky) whereas positive
skewness (a predominance of low outcomes with a few high outcomes) has been
identified with hope and negative skewness (a predominance of high outcomes
with a few low outcomes) has been identified with fear (Hagen, 1969). In this
view preference for the short shot over the long shot would be interpreted as
.. due to the short shot's much lower variance. Likewise, preference for the
riskless lottery over the short shot would be interpreted as a preference for
positive skewness since these lotteries have equal variance.
Moments models have several virtues, not the least of which is that any
distribution can be described, in principle, to any desired level of precision ,
by a sufficiently large set of its moments. But they also have major
difficulties. Some of these are technical as, for instance, 5 the fact that,
subjectively speaking, risk doesn't really act like variance . More serious,
however, is that such theories implicitly assume that moments have independent
psychological reality. That seems doubtful except for the simplest
comparisons. It is not all that easy to intuit the relative variance of
.. lotteries that differ in skewness (e.g, the peaked lottery versus the short
shot) except when the differences are very great (e.g., the peaked lottery
versus the long shot).
In the present theory, lotteries are represented by cumulative graphs
called Lorenz curves that are used in economics to show how wealth is
distributed among people. Welfare economists find them useful for saying
things like "The poorest 20% of the population in Country X have less of their
country's wealth than the poorest 20% in Country Y." What is relevant for us
is that subjects tend to talk as though they also view lotteries in cumulative
terms. Here, for example, are reasons given by three typically risk averse
subjects for why, in a forced choice between the short shot and the long shot,
they prefer the short shot (Lopes, 1986):
I'd rather have greater chances of winning a little something than
greater chances for nothing. The triple jackpot [in the long shot]
doesn't make me want to go for it cuz the odds are too great.
(Subject #10)

I choose the [short shot] because there is only one chance of me J%

 Psychology of Risk / 13

losing and the best odds indicate a good chance of winning $71 or
more. The [long shot] has too many opportunities to lose -- it is
too risky. (Subject #7)

In the [long shot], 32% do better than the best in the [short shot],
but 31% get nothing at all. The [short shot] is the better risk.
(Subject #14)

Notice the inequalities: the keynote of these protocols is the cumulative

likelihood of meeting or exceeding a goal (e.g., "greater chances of winning a
little something," "a good chance of winning $71 or more," "do better than the
best"). The protocols also suggest that the subjects are mostly concerned
about doing badly (getting zero or a small amount).
Here for comparison are protocols from three subjects who chose the long
shot. These subjects were among the most risk seeking of the group. Note
their clear focus on the long shot's large prizes.

The chance for winning nothing is small with the [short shot] but
since the dollar amount in the [long shot] is attractive I run the
risk of losing and go for the [long shot]. (Subject #12)

The top prize money of the [long shot] is better. You still have a
good chance of winning some money in the [long shot] as well as
having a shot at the top prize money. The in between prize money in
the [long shot] is not all that bad, and is greater than the top
prize money of the [short shot]. (Subject #9)
I'll take the added risks of losing it all or getting a lower number
for the chance of the higher prizes. Therefore I'll pick the [long
shot]. (Subject #11)

B. How to Draw a Lorenz Curve (and Miy)

Lorenz curves are convenient for looking at lotteries cumulatively and

for comparing lotteries selectively on either low outcomes or high outcomes.
They also highlight differences and similarities among lotteries that are not
imediately apparent by direct inspection of the lotteries. Figure 3 shows
how a Lorenz curve is graphed. Column 1 at the left shows the prizes for the
peaked lottery ordered from the least (at the top) to the most (at the
bottom). Column 2 gives the number of tickets at each level and column 3
converts these into probabilities. Column 4 is the product of columns 1 and 2
(which is the total prize money at each level). Columns 5 and 6 are running
sums of columns 3 and 4, respectively, and column 7 is column 6 divided by the
total priz? money. This gives the cumulative F-oportion of prize money at
each level .

,.o

"V
 Psychology of Risk / 14

a. Insert Figure 3 about here

The Lorenz curve is plotted at the right. The abscissa gives the
cumulative probability (column 3) and the ordinate gives the cumulative
proportion of prize money (column 7). Notice that the Lorenz curve runs from
the lower left (the low or "bad" end of the curve representing tickets with
small prizes) to the upper right (the high or "good" end of the curve
representing tickets with big prizes). If every ticket in the lottery were a
$100 sure thing, the Lorenz curve would fall on the diagonal. To the extent
that the tickets have unequal prizes, the Lorenz curve bows away from the
diagonal.

Insert Figure 4 about here

Figure 4 gives the Lorenz curves for the long shot and the short shot.
Notice that the curve for the long shot lies everywhere below the curve for
the short shot. This is the sign of large relative dispersion. The long
shot's several large outcomes (indicated by the steepness of the curve at the
upper end) must be paid for by its many zero and small outcomes (indicated by
the flatness of the curve at the lower end). The Lorenz curve for the short
shot, on the other hand, lies nearer the diagonal since it has no really large
outcomes and only a few small outcomes. These considerations lead directly to
a simple rule for choosing between lotteries: people who want to avoid the
wors outcomes should prefer lotteries whose Lorenz curves lie near the
diagonal at the low end (stippled area at lower left), and people who want to
have a go at the best outcomes (at least as good a go as can be gotten) should
prefer lotteries whose Lorenz curves lie far from the diagonal at the high end
(striped area).

Insert Figure 5 about here

Figure 5 shows the Lorenz curves for the short shot and the riskless
lottery. These make an interesting comparison because their Lorenz curves
cross one another: the curve for the riskless lottery is nearer the diagonal
at the low end (stippled area lower left), but further away at the high end
(striped area upper right). Thus, the riskless lottery offers both higher
minima and higher maxima. Not surprisingly, it appeals to both kinds of
subjects. Here are subject #10 (risk averse) and subject #11 (risk seeking)
from Lopes (1986) explaining why they chose the riskless lottery:

The [riskless lottery] has (1) a higher jackpot (2) greater chance
of winning a larger amount under $100. I look at the highest amount
I could lose rather than the highest amount I could win. (Subject
#10)
I picked the [riskless lottery] because both the minimum and the
". maximum amounts are more, and because for both there's a good chance
of getting around $100. (Subject #11)

.1"
r.wJ1 wt . ' . ,. . ' . %_ ' % :q. j. __ 1.I J I . . .. . . . ..... . p [

Psychology of Risk / 15

Insert Figure 6 about here

Figure 6 gives Lorenz curves for the bimodal lottery and the long shot.
These lotteries (which look very different superficially) are similar at their
low ends (lots of small outcomes) but differ markedly at their high ends (lots
of moderately large outcomes versus a few really large outcomes). People who
want to avoid low outcomes should have a mild preference for the bimodal
lottery since it lies a little nearer to the diagonal at the low end. People
who want to win large outcomes, however, should have a relatively strong
preference for the long shot since its Lorenz curve lies quite a bit further
from the diagonal at the high end. This is exactly what happens (Lopes, 1984;
Schneider & Lopes, in press). The pattern is illustrated by the following two
protocols from Lopes (1986), the former expressing a mild preference for the
bimodal lottery and the latter expressing a stronger preference for the long
shot:

I chose the [bimodal lottery], because there seems to be twice as

much chance to get nothing in the [long shot]. Unfortunately,
there's a 50% chance of getting less than $100 in the [bimodal
lottery]. The [long shot] also has higher stakes. However, all
those zeros worry me. (Subject #5)

[Took long shot] because (1)very hi win possible, (2) chance of

winning > $100 about same as for other distribution. [The bimodal
lottery] has too little possible gain for the hi risk of winning
nothing. (Subject #6)

Insert Figure 7 about here

Lorenz curves can also be used to describe loss lotteries. The only
thing that needs to be remembered is that for losses, the best (biggest)
outcome is zero. (Not hard to remember when real losses are being
considered!) Figure 7 gives the Lorenz curves for the bimodal loss lottery
and the long shot loss lottery. Although they look a little different than
Lorenz curves for gain lotteries (the curves now being above the diagonal),
they are read in exactly the same way. Cumulative probability still runs from
0 at the left to 1 at the right and cumulative proportion of value still runs
from the smallest value at the bottom (minus 1) to the largest value at the
top (zero). The worst outcomes are still in the lower left corner and the
best outcomes are still in the upper right corner. If all the tickets were
for a $100 loss, the Lorenz curve would fall on the diagonal.

The rule for choosing also stays the same: people who want to avoid the
worst outcomes (big losses) should prefer lotteries whose Lorenz curves lie
near the diagonal at the low (bad) end (stippled area); people who want to
obtain the best outcomes (small losses) should prefer lotteries whose Lorenz
curves lie far from the diagonal at the high (good) end (striped area).

These different reasoning patterns can be seen clearly in subjects'

protocols. Here for example are two typically risk averse subjects from Lopes
(1986) explaining why they prefer the binoctal loss lottery to the long shot

~ ~
.b . . .. . .q ~. .4 A . . . .. .. . . . . . . . .. .. . . . . . . . . . . .. --
 Psychology of Risk / 16

loss lottery:
With the [bimodal lottery], the most that I can lose is $200. With
the [long shot], I could lose $439. (Subject #2)
I would not risk losing $439, or even $292 and up. (Subject #3)
In contrast, here is a subject who chooses the long shot:
I choose the [long shot] because there is a preponderance of tickets
that can incur no loss, and a fair number of other tickets that
could lose less than $98. In the [bimodal lottery] 50% of the
tickets do stand to lose $93 or less but there are fewer that can
promise to cause no financial loss. I notice that there are large
amounts to be lost if one is unlucky, but the chances of being
unlucky are somewhat slimer in the [long shot]. (Subject #1)

Negative lotteries are particularly interesting because they often

present difficult choices. Here, for example, is a subject who has chosen the
peaked loss lottery over the rectangular loss lottery:
I go back and forth on this, the gain on improving the chances on a
low loss increases the chance of a higher loss. I pick the [peaked
lottery] to try to reduce the higher loss. (Subject #11)
Another subject, however, chooses to gamble on the long shot rather than take
a sure $100 loss:
In the [sure thing] no way could I lose more than $100. But no way
could I lose less, either. In the [long shot] there are enough
chances to lose less than $100 to justify losing a lot more. I'll
go ahead and see if I can get less than $100 loss -- maybe even zero
loss -- rather than accept a sure loss of $100. If I lose $100 I
might as well lose $439. I don't like it either way. However, the
[chances] of losing $439 or $390 or $341 -- etc. down to $146 are
quite high. But so are zero, $49, or $98 or $146. So I'll take the
chance. I don't feel great about it, though. (Subject #13)
Conflicts such as these are relatively common for loss lotteries. This is
important because it suggests the existence of an additional factor in risky
choice that is not captured by the Lorenz curve analysis. This factor will be
discussed below.

Insert Figure 8 about here

Finally, Lorenz curves for loss lotteries can also cross one another
producing agreement in choice between risk averse and risk seeking
individuals. Figure 8 shows the Lorenz curves for the riskless loss lottery
and the short shot loss lottery. Below are protocols from the same two
subjects (the former risk averse and the latter risk seeking) who previously
defended their choice of the riskless lottery over the short shot for gains.
Below they tell why they reject the riskless lottery for losses:

- --..- "~
" .,.,";b
--i'-i': . .'" ":" -," .""-* . "........ v ....... ; ...2'> ...... > .' .i.;,....;.i . ..... .
 Psychology of Risk / 17

[Took short shot because] (1) there is a greater chance of losing

$130 or more with the [riskless lottery]. (2) There is a greater
chance of losing less than $70 in the [short shot]. (Subject #10)

I pick the [short shot] because the maximum loss is less and because
you may be able to hit as low as zero loss. No matter what, you
lose $70 and possibly $200 in the [riskless lottery] -- too much
risk. (Subject #11)

To sum up, Lorenz curves have at least four virtues for representing
lotteries. First, being cumulative, they reflect the fact that subjects tend
to evaluate lotteries in terms of inequalities. Second, they facilitate
comparison of the particular regions in lotteries that appear to be salient to
people with different goals. Third, they suggest similarities and differences
among lotteries that, although they may seem obvious in retrospect, do not
easily come to mind from inspection of the lotteries themselves. Fourth, they
predict when people having different goals will agree (or disagree) about
lotteries and when differences will be relatively large (or small). These are
no small virtues for a stimulus representation. But that is all that Lorenz
curves are: a way of representing lotteries. It is not to be supposed that
people convert risks into "mental Lorenz curves". Clearly, they do not. The
purpose of the Lorenz curve representation is to deepen understanding of the
functional stimulus and to help us psychologists see (literally) what it is
that people focus on when they compare lotteries.

IV. A Two-Factor Theory for Risky Choioe

Until fairly recently, personologists focused largely on variables inside

the skin, leaving experimentalists to deal with situational variables.
Lately, however, personologists have come to agree with Mischel's judgment
that "traditional trait-state conceptualizations of personality, while often
paying lip service to man's complexity and to the uniqueness of each person,
in fact lead to a grossly oversimplified view that misses both the richness
and the uniqueness of individual lives" (Mischel, 1968, p. 301).
Nevertheless, it is clear that "while stimuli or situations come to evoke and
maintain behavior patterns, they do not respond by themselves" (Mischel, 1968,
p 295). Behavior has sources both inner and outer. We are disposed by our
unique constitutions and histories to behave in certain ways, but it is
"situational stimuli that evoke [our responses], and it is changes in
conditions that alter them" (ischel, p. 296).

Two-factor theory uses both a dispositional factor and a situational

factor to explain risky choice. The dispositional factor describes the
underlying motives that dispose people to be generally oriented to achieving
security (i.e., risk averse in conventional terminology) or to exploiting
potential (i.e., risk seeking in conventional terminology). The situational
factor describes people's responses to inediate needs and opportunities. As
will be shown, these factors are sometimes in conflict and sometimes in
concert, producing complex patterns of behavior in which risk averse choices
 Psychology of Risk / 18

and risk seeking choices exist side by side in the same individual's behavior.

A. Factor 1: Security versus Potential

In an earlier article (Lopes, 1984), I argued that risk averse and risk
seeking individuals differ in whether they pay most attention to the worst
outcomes in a distribution or the best outcomes. Risk averse people appear to
be motivated by a desire for security whereas risk seeking people appear to be
motivated by a desire for potential. The former motive values safety and the
latter, opportunity.

In mathematical terms, security motivation corresponds to weighting the

worst outcomes in a lottery more heavily than the best outcomes and potential
motivation corresponds to the opposite. Such processes could be modeled
mathematically by applying appropriate weights to the cumulative functions
that constitute the data for Lorenz curves. Indeed, there are several
functions currently in use by welfare economists that could be used to
quantify risk averse and risk seeking preferences (A.Atkinson, 1970; Dahlby,
1985). Describing these functions will not be necessary for present purposes,
however, since raw Lorenz curves show all we need to know.

There are, however, two general points that should be made about
weighting. First, weights in the theory are joint functions of the magnitudes
of probabilities and the magnitudes of the outcomes to which they are
attached. This is a fundamental departure from the family of weighted value
models since in those models, probability and value are independent. Second,
weights reflect individuals' Qoals and not their perception of probabilities
or values. Thus, the fact that a person chooses, for example, to minimize the
likelihood of a bad outcome does not imply either that (subjectively) he
underestimates the value of good outcomes or that he overestimates the
probability of bad outcomes. Although psychophysical effects may occur in
either the money or the probability domain, these are considered to be of
secondary importance in determining risky choice.
The security/potential factor is conceived to be a dispositional
variable, reflecting the way individuals typically respond to risks. Not
surprisingly, security motivation (risk aversion) is the far more connon
pattern (see Lopes, 1984; Schneider & Lopes, in press), so cormon, in fact,
that economists have considered it to be the pattern for Everyman (Arrow,
1971; Pratt, 1964). This is probably not due to chance. Standards of
prudence are passed from parent to child in the normal course of growing up.
If that is not enough, hard experience informs us in no uncertain terms that,
as Damon Runyan said, "the race is not always to the swift nor the battle to
the strong, but that's the way to bet" (cited in Ellsberg, 1961, p. 644).
Risk seekers, on the other hand, may dog the long shots, waiting (as a famous
risk seeker once said) for "that one streak of luck, properly ridden and
encouraged," to compensate them for all the bad times (Thackrey, 1968, p. 67,
quoting Nick the Greek).

It should also be noted that the fact that someone is primarily motivated
by one of the poles of the security/potential dimension does not imply that
they are unaware of the other pole. It is better to think of these opposing
tendencies as existing in some strength in everyone (as do M s and Maf in

7W
 Psychology of Risk / 19

Atkinson's, 1983, theory), but with potential much less important than
security and aspiration for risk averse people and security much less
important than potential and aspiration for risk seeking people. Such weak
motives would tend to come into play primarily when the stronger motives were
insufficient to determine choice.

B. Factor 2: Aspiratic Level

The security/potential factor reflects the way that a person usually

looks at risks. Risk averse people look more at the downside and risk seekers
more at the upside. But risk seekers may play it safe from time to time, and
even the most risk averse person will take chances -- even big chances -- when
necessary. Aspiration level (Lopes, 1983; Siegel, 1957; Simon, 1955) is a
situational variable that reflects the opportunities at hand ("What can I
get?") as well as the constraints imposed by the environment ("What do I
need?")

The aspiration level that functions in any given situation (including the
present task situation) can reflect at least three different sources. The
first is the direct assessment of what is reasonable or safe to hope for. For
illustration, here is a subject from Lopes (1986) who has rejected the short
shot in favor of the peaked lottery:

The chances are in the [peaked] lottery that I will get something
close to $100, and I might get much more. I don't know why [I
should] let $130 be the top limit when there's a reasonable chance
of nearing $100 and a possibility for more. (Subject #3)

The next subject has chosen the riskless lottery in preference to a $100 sure
thing:

Since I am assured of winning something I am willing to risk a

moderate amount for the possibility of a substantially greater
amount. (Subject #4)

In both these cases, the subjects have taken the riskier option, but not
before assuring themselves that it's prudent to shoot for its somewhat higher
prizes.

The second source of aspiration levels is the direct contextual influence

of the other alternatives in the choice set. Here, for example, is a subject
choosing between the short shot and the riskless lottery:

I chose the [riskless lottery] because I am assured of winning at

least $70. In addition, I have a better than even chance of winning
more than $70. It is the assurance of winning $70 that appeals to
me. (Subject #7)

Based on this rationale, the subject's aspiration level appears to be no

higher than $70. However, when the same subject is given a choice between a
sure $100 (the sure thing) and the riskless lottery she says,
L Psychology of Risk / 20

I chose the [sure thing] because I would rather take the $100 as a
sure thing than risk winning less. The other lottery also offers a
sure thing ($70 at the least), however, the chances of winning less
than $100 are about 50-50 in that lottery, so I opt for the safe bet
of $100, a sure thing. (Subject #7)

The same shifting of aspiration level also occurs for losses. Here are
two more protocols from the previous subject. In the first she rejects the
long shot loss lottery in favor of the short shot. In the second she accepts
the long shot in favor of a sure $100 loss:

I chose the [short shot] because the most I could lose would be $130
and that seems safer than the [long shot]. Also the odds in both
lotteries seem to favor a loss of between $50-$150, so I figure the
lottery which has the lowest ceiling on a possible loss is the
safest risk. (Subject #7)

A $100 loss up front is too hard for me to swallow -- I chose the

[long shot] as it allows for many chances to lose less than $10L.
True, the maximum loss could be as high as $439, but it is still a
risk I am willing to take. (Subject #7)

Notice that the subject seems to switch from considering a $130 loss to be
acceptable to considering a $100 loss to be unacceptable. Statements like
these make it clear that sure things have a powerful influence in organizing
choice, and the same seems to be true of values that are highly likely, though
not certain.

The seemingly special status of certainty in risky choice has received a

prominent role in several th.ories (Allais, 1979; Kahneman & Tversky, 1979;
Machina, 1982), but the mechanism through which certainty effects operate is
as yet unclear. One possibility (Kahneman & Tversky, 1979) is that they are
instances of subjective category-boundary effects in the perception of
probability (see Section II). Another possibility, however, is that certainty
is objectively special since it permits planning to proceed uninpeded by
uncertainty about outcomes that may not be resolved in the near future (see
Section VI.C).
The third way that aspiration levels get set is by outside influence. A
study currently in progress (Lopes & Casey, 1986) is looking at the role of
necessity in a risk-taking task. In the task, subjects play a competitive
game involving multioutcome lotteries against an opponent (either a computer
or another real subject). In the game, players attempt to take or defend
territory on a game board by choosing among moves that have distributions of
possible outcomes similar to the lotteries in Figure 2. Although the data are
not completely analyzed, we have noted a tendency for subjects to prefer
riskier moves when they are in a bad position near the end of a round. This
is as it should be: if there is little or no chance that the safer option
will yield sufficient territory for a win within the number of moves
remaining, the riskier option may be thj probabilistically best choice (i.e.,
more likely to yield a winning outcome)'.

',°% '
 Psychology of Risk / 21

Finally, it should be noted that although aspiration level is

situational, it probably interacts with the security/potential factor, with
security motivated people tending to set more modest aspiration levels than
potential motivated people for both gains and losses. This possible
interaction necessarily complicates the independent assessment of the
contributions of security/potential and aspiration to risky choice.
Nevertheless, support for the conceptual distinction between the two factors
exists in the fact that, as will be seen, the factors often act in opposition
to one another.

Conflict Betwim Security/Potential and Aspiratio"

A simple truth: you can't have it all. Corollary: you want more than i%
you can have. Conflict is in the nature of things. All resources are limited
(wealth, youth, even free lottery tickets), so people get plenty of practice
juggling inconsistent desires and jury-rigging tolerable compromises.

Conflict arises in two places in the present theory. One, obviously, is

the conflict between security and potential. It is a truism in the investment
world that risk and return go together. If you want safety, you pay for it in
yield; if you want yield you pay for it in worry. To say that
security/potential defines a dispositional variable is to say that people
typically choose one way or the other between avoiding bad outcomes and
approaching good outcomes. But this does not mean that people do not see what
they do not choose. In making a clear decision for, say, security, a person
may acknowledge regretfully the loss of opportunity. People also are quick to
notice the special benefits of choices 8 such as the riskless lottery that allow
them to have their cake and eat it too .

The second form of conflict is both more interesting and less obvious.
These are the conflicts that can be created as different situations induce
different patterns of agreement and disagreement between dispositional motives
toward security or potential and the immediate needs and opportunities
affecting aspiration level.

Consider someone who is dispositionally motivated to achieve security and

suppose that, in the present task situation, the person has a modest
aspiration level, say $50. Faced with the choice between the short shot and
the long shot, the person would tend to reject the long shot on both counts:
it is clearly less secure in Lorenz curve terms and it is also less likely to
satisfy the aspiration level. The same would be true for almost any pair of:
gain lotteries. This is because there is a positive correlation between the
ordering of the lotteries in terms of security and the ordering of the
lotteries in terms of the probability that they will achieve the aspiration
level.

For losses, however, there is a conflict between security and aspiration.

Consider the same person choosing between the same two lotteries, but this
time for losses, and suppose that the aspiration level is to lose no more than
$50. The short shot is obviously more secure since its losses are capped at
$130, but it is much less likely to yield a loss of $50 or less. This would

4.]
 Psychology of Risk / 22

be true for almost any pair of loss lotteries: the ordering on security runs
essentially opposite to the ordering on aspiration level.

For a potential motivated person, the situation would be just reversed.

For losses, potential and aspiration level are positively correlated but for
gains they are quite likely to be negatively correlated.

Conflict between security/potential and aspiration can produce quite

complex patterns of data (see Coombs & Avrunin, 1983, for a general discussion
of data patterns produced by conflict). Table 1 gives some values for
illustration. The top of the table is for a risk averse individual and the
bottom for a risk seeker. Gain choices are on the left and loss choices on
the right. The aspiration level of the risk averse person is assumed to be
$50 for both gains and losses whereas the aspiration level of the risk seeker
is assumed to be $80 for gains and $20 for losses. These values are purely
hypothetical, but they accord with our intuition that risk averse people
probably have more modest aspiration levels than risk seekers.

Insert Table 1 about here

Let's begin with the risk averse person. Column 1 lists the six
lotteries from Figure 1 plus a $100 sure thing. Columns 2 and 6 (AL) give the
probabilities that the lotteries will yield the aspiration level: $50 or more
for gains and $50 or less for losses. Columns 3 and 7 (SEC) give hypothetical
values on security. (Keep in mind that the riskless lottery and the short
shot change places as one goes from gains to losses.) These values range
evenly between 1 for the sure thing and 0 for the long shot except for a tie
in each ordering. (For gains, the worst outcomes in the short shot and the
peaked lottery are almost identical in probability and value. Their Lorenz
curves would be essentially superimposed at the low end. In the same way, for
losses, the worst outcomes for the riskless lottery and the peaked lottery are
almost identical. Their Lorenz curves would also be superimposed at the low
end.)
Columns 4 and 8 (SEC x AL) show how security and aspiration are
integrated. A multiplying rule is used because the choice is conjunctive. A
lottery lacking good values on either security or aspiration will be rejected.
The final two columns (REL PREF) simply normalize the products to a cormon
base (by dividing by the sum of products) in order to allow an easier
comparison of relative preference.

A similar analysis is given in the bottom of the table for the risk
seeking person. Values for potential (POT) have replaced values for security,
and the aspiration levels are now $80 or more for gains and $20 or less for
losses. Note here that the ties in potential are now between the riskless
lottery and the peaked lottery for gains and the short shot and the peaked
lo"tery for losses.

Starting first with the risk averse person, note that relative
preferences for gain lotteries tend to decrease from the sure thing to the
long shot. This reflects the strong positive correlation between SEC and AL
for gains (r = .97). For losses, however, there is an inverse-U pattern:
preferences are low at the extremes but higher in the middle. This reflects

. .. . . . . .. . . . . . . . . . . . . .
 Psychology of Risk / 23

the strong negative correlation between AL and SEC for losses Lr = -.91).

For the risk seeker, the simple pattern occurs for losses: preferences
tend to increase from the sure thing to the long shot reflecting the strong
positive correlation between POT and AL for losses (r = .92). For gains,
however, the pattern is complex. The least preferred lotteries are the sure
thing and the short shot and the most preferred are the bimodal, rectangular,
and the long shot. This complexity reflects the negative correlation between
POT and AL for gains Lr = -.97).
I"

To sum up, the two-factor theory integrates a dispositional tendency to

seek either security or potential with situationally driven aspiration levels.
Security motivation captures the Bernoullian (1967) intuition that people are
generally disposed to prefer sure things and gambles without large chances of
bad outcomes. However, the theory handles equally directly the less prevalent
tendency of some people to approach long shots and other gambles offering the
unlikely possibility of large outcomes. In addition, the theory deals
directly with situational circumstances that may cause a person to experience
conflict between dispositionally driven preferences and externally driven
goals. Thus, the theory explains how the person can be risk averse in the
economic sense (i.e., typically preferring sure things) but sometimes make the
same choices as someone who is ordinarily risk seeking.

V. Evidence for the Two-Factor Theory

A. Riskiness is the Absence of Security

The first bit of support for the two factor theory comes from judgments
of riskiness. In Experiments 3 and 4 of Lopes (1984), subjects were shown
pairs of gain lotteries and were asked to say which was the riskier. In
virtually every case, the lottery judged to be the riskier was the one whose
Lorenz curve lay further from the diagonal at the low end. The only
exceptions involved the relative riskiness of riskless lotteries and short
shots. (There were three examples of each). About half the subjects judged
the riskless lotteries to be the riskier (contrary to the original
expectation) and about half judged the opposite.

Experiments 5 and 6 of the same study suggested why this was so. in the
former experiments, the term "risk" was left vague so that subjects could
supply their own meanings. In the latter experiments, however, subjects were
asked to select the lottery for which it would be riskier to pay $100. Under
this condition, judgments for the other lotteries were virtually unchanged,
but subjects were now nearly unanimous that the riskless lotteries were the
riskier, a judgment that makes objective sense because there is a good chance
that riskless lotteries will yield substantially less than $100 (e.g., $70).
Apparently in the original experiment subjects adopted different aspiration
levels. For most lotteries, riskiness does not depend on whether the
aspiration level is low ($50) or high ($100), but for the riskless lotteries,
the shift in aspiration is crucial.
i
The ability of the two factor theory to account for judgments of
riskiness is a point in its favor, particularly as contrasted with

..................... ...........
 Psychology of Risk / 24

psychophysical models. In the latter models, there is no such thing as risk.

Although they predict risky choice, they are silent on judgments of risk.
Intuitively, however, risk plays a role in risky choice. Risk is the absence
of security; security is the absence of risk. Seems simple enough.

B. Risk Attitude Is More Than the Psychophysics of toey

The second bit of support for the theory is that it can predict the
preferences of both risk averse and risk seeking people. In experiments 1 and
2 of Lopes (1984), subjects were shown various pairs of lotteries and asked to
say which they would prefer to play. The subjects were then divided according
to whether or not they tended to take the sure thing when it was offered.
Risk averse subjects (i.e., those subjects who took the sure thing 8 or more
times out of 10) had preferences that were essentially perfectly predicted by
security motivation (i.e., they preferred lotteries whose Lorenz curves lay
near the diagonal at the low end.) Risk seeking subjects (i.e., those who
took the sure thing 3 or fewer times out of 10) had preferences that were for
the most part predicted by potential motivation (i.e., they preferred
lotteries whose Lorenz curves lay far from the diagonal at the high end).

The ability to account for people whose choices are primarily risk
seeking is another benefit of the two factor theory. Psychophysical theories
and moments theories are theories of Everyman because they are based
mechanistically on principles that should hold for us all: "our perceptual
apparatus is attuned to the evaluation of changes or differences" (Kahneman &
Tversky, 1979, p. 278); "uncertainty ... has a disutility growing worse with
increasing speed when [the] standard deviation [of utilities] increases"
(Hagen, 1979, p. 274). But-Everyman is risk averse for gains even though
every man (or woman) is noty. Two factor theory puts risk seekers and risk
averse people on equal footing. Although their choices may differ profoundly,
their choice processes have more similarities than differences. They
understand risks in the same way (cumulatively) and they trade off the same
factors. Their goals may differ, but they have the same conceptual equipment.

C. Risky Choice Is Not Conflict Free

The best evidence for the two factor theory comes from a recent study of
the preferences of preselected risk averse and risk seeking subjects for gain
and loss lotteries (Schneider & Lopes, in press). Subjects were selected from
a large group of undergraduates who had filled out a brief questionaire asking
for their preferences in five choice pairs. Each pair contained a positive
two-outcome gamble and a sure thing of equal expected value. In accord with
conventional usage, risk averse subjects were defined as those who selected
the sure thing every time, and risk seeking subjects were defined as those who
selected the gamble at least four times. Thirty subjects were selected from
each group.

The 10 stimuli in the experiment included the 6 stimuli in Figure 1 plus

a $100 sure thing. Subjects were given the stimuli in all possible pairs and
asked for their preferences. The pooled data are in Figures 8 and 9 for risk
averse and risk seeking subjects, respectively. The stimuli from Figure 1 and
the sure thing are listed on the abscissa by letter codes (ST = sure thing, RL
riskless, SS = short shot, PK = peaked, RC = rectangular, BM = bimodal, LS =
 '.t-. .- -. -.. . .- k , . ' - ) y M.< V u

Psychology of Risk / 25

,S

long shot). The lotteries identified by number were an additional riskless

lottery (2), an additional short shot (4), and an additional long shot (8).
The open symbols are for gain lotteries and the filled symbols are for loss
lotteries.

Insert Figures 9 and 10 about here

Looking first at the risk averse subjects, it is clear that their

preferences for gains decrease essentially monotonically from the sure thing
to the long shot. Their preferences for losses, however, have an inverse-U
pattern, being low for the sure thing and the riskless lotteries (at the left)
as well as for the long shots (at the right). These qualitative patterns
(which were highly reliable statistically) are exactly what the two factor
theory predicts (cf. Table 1). For gain lotteries, security and aspiration
support the same choices, producing monotonicity. For losses, however,
security and aspiration conflict, producing nonmonotonicity.

The risk seeking subjects also follow the pattern predicted by the two
factor theory. Their preferences are essentially monotonically increasing
from the sure thing to the long shot for losses, but vary complexly for gains.
Worst liked are the sure thing, the short shots, and the peaked lotteries;
long shots and riskless lotteries are better liked. These patterns were also
highly reliable statistically.

Considering the crudeness of the estimates used in Table 1, the fit is

remarkably good. Both groups like RL a little better than they should for
gains, and risk seekers like long shots a little less well than they should
for losses, but the overall picture is as it should be. The presence of
nonmonotonicity in both data sets confirms the existence of conflicting
factors in risky choice. And the mirror-symmetry between the two subject
groups confirms their structural similarity. They are basically doing the
same thing, but their values differ at least on security/potential and
probably on aspiration level as well.

VI. The Things We Don't Talk About

Psychologists who study risky choice don't talk about a surprisingly

large number of factors that are psychologically relevant in choosing among
risks. Given the poverty of our theoretical language, it is, perhaps,
surprising that we manage to talk at all. Here are some words that are not to
be found in the theoretical vocabulary: fear, hope, safety, danger, fun,
plan, conflict, time, duty, custom. Nor can these words be given meaning in
psychophysical theories. The language of psychophysics is the language of
perceptual distortion. It has room neither for the experience of emotion
(fear, hope, fun), nor for the physical context in which risks occur (safety,
danger), nor for the social constraints on individual action (custom, duty),
nor for the cognitive activity of the chooser (plan, conflict).

Distributional theories, on the other hand, can express these meanings

easily. Oddly, however, it was an economist (Hagen, 1969) who identified
skewness with hope and fear whereas psychologists working with similar

-i ~ ~ ~ ~ ~ - - '*
 Psychology of Risk / 26

theories (Coombs, 1975; Coombs & Lehner, 1981; Luce, 1980; Pollatsek &
Tversky, 1970) have shied away from using such language. Queasiness about the
ordinary language of emotion and intention goes back in psychology at least to
Watson's behaviorist manifesto, and in the area of risky choice has been
amplified by exposure to a similar movement in economics termed positive
economics (Friedman, 1953). In fact, modern economic expected utility theory
(von Neumann and Morgenstern, 1947) has relegated even the classical
psychophysical concept of utility to the status of an epiphenomenon. In the
modern view, utility does not precede and cause preferences; it is, instead,
merely a convenient fiction that can be used by the practitioner to summarize
the preferences of those who, by choice or chance, follow the dictates of the
von Neumann and Morgenstern axiom system.

If however, hope and fear and plans are necessary ingredients in risky
choice, then it is not unscientific to talk about them. Many social
scientists working outside the narrow confines of the laboratory or the
mathematical proof treat these terms casually, like old friends. In the
remainder of this paper I draw on these broader approaches to risk and show
their relation to the two factor theory.

A. Fear and the Safety-First Principle

Psychophysical theories of risky choice do without a psychological

concept of risk, but people cannot. Risk is a fact of our physical and social
environments. This is nowhere more clear than in agriculture, particularly
subsistance agriculture in which one's livelihood can be literally threatened
from all sides (by floods, by pests, by invading armies). Agricultural
economists look at risk quite differently than either psychologists or those
economists who have influenced psychology most strongly in recent years. For
the agricultural economist, both risk and risk aversion are real. The
behavior is connected to the phenomenon by a simple rule called the safety-
first principle.

The subsistance farmer is in a difficult position. Food crops provide L

food for the table and have low variance of return, but their expected return
is also low. Cash crops are more variable but have higher expected return.
The problem is how much of each crop to plant. Although many different models
of farmers' choice processes have been developed (Anderson, 1979), they all
boil down to a simple rule: first take care of subsistance needs (food for
the larder and seed for the coming season) and then plant cash crops.
Although the farmer's world may be risky, the farmer does not view himself as
gambling. In Ortiz's (1979) words, "The peasant's preference for subsistence
over starvation cannot be rephrased into a preference for X chance of Y income
over X-n chance of Y-m income; such paraphrasing totally misrepresents his
options and it is unnecessary" (p. 235).

It should also be noted that although fear feeds risk aversion, risk-
taking can be fed either by hope or by necessity (see Section IV.B).
Conventional economic theory treats risk-taking as a luxury. Neither
individuals nor firms are supposed to indulge in it unless they can afford it.
Kunreuther and Wright (1979), however, have pointed out that sometimes the
poorest farmers devote as large a proportion of their land to cash crops as
the richest farmers. Likewise, Bowman (1982) has demonstrated that

Lt'
 . •. . .L . . ,, . . , . %: . ,. ,. l.-i , 6;L , ,-v-- ,T _ ,VT :_-V.rV :,:. VT _

Psychology of Risk / 27

economically troubled firms often engage in riskier behavior than economically

sound firms. In both these cases, risk seeking behavior is interpreted as
arising from necessity. If there is not enough safety even when safety is put
first, the risky choice may be the only choice.

The connections between safety-first and the two-factor theory are

obvious. The safety-first principle aims at security. A target level is set
and choices are made so as to maximize the probability of achieving the
target. The resulting choices are "cautiously suboptimal" (Day, 1979).
However, risk-taking behavior may predominate when the aspiration level cannot
be achieved safely. In neither case do the choices reflect the psychophysics
of money or of chances. Nor are they merely the signs of pessimistic or
optimistic world views. Farmers who follow the principle want (and need) to
accept the risks of cash cropping, but they can do so only up to a point.
Their choices reflect the planful activities of intelligent, though
unsophisticated, people who know the odds and bet that way.

B. Planning Is Applied Hoping

The flip side of safety-first is entrepreneurship in which safety takes
second place to opportunity. McClelland (1961) described entrepreneurs as
people who display a willingness to take risks in situations in which their
skills and effort can make a difference. The issue of control appears to be
particularly important. Although spectators may judge that entrepreneurs take
more risks than the average person, entrepreneurs see themselves as not being
particularly risk prone (Keyes, 1985, pp.207-209; McClelland, 1961, p.2 2 2 ).
This is also true of people who engage in physically risky professions or
hobbies: "First they challenge fate, and then they try to bullwhip fate to
its knees by making their adventure as predictable as possible. In fact, it's
not adventure they're after. It's mastery" (Keyes, 1985, p. 115). Even in
casino gambling where outcomes cannot be controlled, high stakes gamblers take
pride in the control of their emotions (see for example, Alvarez, 1983,
pp. 47, 169; Thackrey, 1968, pp. 62-63).

A lot has been written about control in recent years. Langer (1977) has
synthesized an impressive body of experimental evidence documenting the fact
that people behave as though they believe that chance events can be
controlled. In her view, the illusion of control comes about for several good
reasons: (1) people are motivated to master their environment, (2) it is
unpleasant to believe that one has no control, (3) chance and skill elements
coexist in many situations, and (4) the illusion can help us emotionally more
than it can harm us practically.

One might suppose (as Langer has not) that risk-taking behavior is caused
by the optimistic illusion that outcomes are more controllable than they
really are. McClelland (1961), however, has argued that entrepreneurship is
independent of optimism. He presents evidence from four different countries
that school boys who are high in optimism are "conscientious, efficient,
forward-looking, manaQerial types [who work] hard and efficiently at
everything more or less indiscriminately" (p. 227-228), but they are not
necessarily high in achievement motivation. A student who is high in
achievement motivation works hard only on those things that can give a sense
of personal responsibility. "If there is no challenge, he doesn't work so
*,9
 - 1 w...ILI

Psychology of Risk / 28

I
hard: in this sense he would make a poor bureaucrat" (p. 228).

The belief that one can control one's fate appears to be necessary to
good mental health (Abramson, Seligman, & Teasdale, 1978). When highly
desired outcomes are believed to be unlikely or when highly undesirable
outcomes are believed to be likely, and when the individual believes that
nothing can be done to change these likelihoods, depression results causing
attendant motivational and affective deficits.

Beliefs about control and motivation feed back on one another. When
aversive events occur despite one's efforts to prevent them, motivation to
control events is reduced both in animals (Maier & Seligman, 1966) and in
humans (Hiroto, 1974; Hiroto & Seligman, 1975). When control is later made
possible, the motivational deficit prevents learning from occurring.
Responses that are not made cannot be reinforced. A vicious cycle results.

Feedback between control and motivation can also be positive. Sawyer

(cited in MClelland, 1961, p. 222) has argued that self-confidence in
individuals can lead to the very circumstances that are necessary to achieve
success. He gives as an example the settling of the American West in the 19th
century. Had not so many people behaved as though it was possible, it would
not have been possible. Although individuals failed in large numbers,
sufficiently many survived that settlement was collectively accoplished.

In the two-factor theory, people may be motivated by achievement

(potential) as well as by security. McClelland (1961) stressed in his work
that people who are high in achievement motivation do not like to gamble.
That may be so in the narrow confines of two-outcome gambles presented in the
laboratory. In the real world it seems unlikely that achievement motivated
people could be obsessed by security. Although one cannot control the actual
outcome in chance situations, one can control the likely outcome. This is
what portfolio managers do in the investment field. Maximum return and
maximum safety cannot both be had, but one can assemble portfolios that trade
a little risk for an acceptable return. Balancing the unavoidable risks in
one's personal portfolio is a skillful activity that should appeal to
achievement motivated individuals.

C. and

Uncertainty is embedded in time. There is a now in wh. 2h some tn.:ngs are ,"
true, a future in which other things may be true, and a still farther future
in which we may reflect on the past. At the point of choice we look forward
along this track, and we also anticipate looking back. T7he temporal element
is what gives risk both savor and sting.

Pope (1983) has criticized expected utility theory for having i'ncred the
time between the decision and the resolution of the uncer-tainty, what she
calls the pre-outcome period during which fear and hope operate. She points
out that the period can be long not only for long-term decisions such as
individuals choosing careers, governments embarking on social programs, and
businesses making major capital investments, but also for repeat short-term.
decisions: "after deciding to devote a fraction of the housekeeping funds to

i ',
 Psychology of Risk / 29

a weekly lottery ticket, housekeepers can dream from age nineteen to ninety-
nine that they will become millionaires after the next drawing" (p. 156).

Uncertainty in the pre-outcome period can be pleasurable or

unpleasurable. A traveler to Las Vegas may pay an extra fee to avoid the
worrisome uncertainty of the $1,000 deductible on a rental car and then
proceed at all speed to a casino in order to purchase the delightful
uncertainty of gambling. Insurance cannot stop disaster from happening, but
it can stop worry. Gambling cannot guarantee future bliss, but it can give
hope. For small amounts of money we can enjoy the current psychological
benefits of either in the same way that we enjoy other psychological
commodities such as entertainment and convenience.

Decisions involving larger amounts of money and other sorts of serious

consequences require us to extend our analysis beyond the time in which the
uncertainty is resolved. Fears and hopes are then no longer relevant, but
regret and disapointment are. Bell (1982, 1985) has argued that if we
experience regret over decisions that turned out badly or disappointment over
outcomes that fail to match up to expectations, these factors are as important
to consider as more tangible monetary benefits and losses. "Psychological
satisfaction, as opposed to the satisfaction derived from consumption, is an
appropriate objective that should be included in any decision analysis if the
decision maker regards it as a criterion for decision" (Bell, 1985, p. 26).

Decision makers also look to a future in which their preferences may

change (March, 1978) and in which options not currently imagined may have
become possible. Day (1979) has pointed out that farmers and business
managers alike resist procedural changes that put them too far from current
practice. This is a sensible rule for adapting to dynamic environments (both
internal and external) since it leaves the old practice available as a
fallback position should the new policy not produce the expected effects.

The importance of having safe fallback positions in real life may account
for the fact that the perceived riskiness of technological hazards is not
solely related to estimates of annual fatalities but also reflects dread of
outcomes that are perceived to be uncontrollable, catastrophic, not easily
reversed, and of high risk to future generations. Thus, intelligent but
technically unsophisticated raters (students and members of the League of
Women Voters) estimate fewer annual deaths from nuclear power than from home
appliances but nevertheless consider nuclear power to be much more risky
(Slovic, Fischhoff, & Lichtenstein, 1980, Tables 2 and 3).

Psychophysical theories of risk do not consider time, although one might

conceive a psychophysics of future events. Economists use a similar notion
(time preference rates) to handle the fact that money now is generally worth
more than money later. But as Pope (1979) has argued, such a concept would
deal only with changes in the worth of the final outcomes and not with the
emotions that one experiences in the interim. Theories (such as the present
theory) that recognize planning and conflict resolution as an integral part of
risky choice can deal more naturally with such temporal factors both as they
affect changes in the aspiration level and as they affect position on the
security/potential dimension.

" -
-. " , ,. ,' b-
 Psychology of Risk / 30

D. Shezpas and Otber High Rollers

All life chooses among risks, though we do not ordinarily think of trees
choosing how to gamble their seed or amoebas choosing whether to approach or
avoid possible prey. For the lowest organisms, evolution has done the
choosing and equipped them with prewired choices. For higher organisms,
however, cognition increasingly intervenes, allowing learning and reason to
override rote instinct. Humans are the most complex cognitively and exhibit
responses to risk that sometimes have little to do with the satisfaction of
immediate wants or needs. Among these are the responses of our social selves.

High standard mountaineering is incredibly risky. The chances of being

killed on a Himalayan climb are about 1 in 10. Why is it done? What is there
on the top of such mountains that anyone might want? Obviously, the answer is
aesthetic, not practical, at least for the recreational climber (if such a
word can be applied to so hazardous an avocation) who must expend considerable
personal resources and obtain even more considerable institutional resources
just to make the attempt.

But the Sherpas who carry the loads also share in the risks. Why do the
Sherpas climb? The conventional view is that they climb out of economic
necessity. Michael Thompson (1980), however, himself an Everest climber,
disputes this distinction between Sherpa and non-Sherpa. In Thompson's view,
risks that are pursued for practical purposes become tame in the process as
has commercial air travel. But risks that are pursued for themselves do not
become tame. Thus, for aesthetic purposes a proposed Everest route "is only
felt to be worthwhile if there is considerable uncertainty as to its outcome"
(p. 278). Sherpas also take this view and refer scornfully to the route that
Hillary and Tenzing followed, the easiest of the routes for obvious reasons,
as "the Yak route." A small joke among those who share a common aesthetic.

Risk taking is one of the ways we define ourselves psychologically and

socially (Douglas and Wildavsky, 1982). It is a mistake to suppose even in
the realm of financial risks that choice is a purely monetary matter. For
entrepreneurs and high stakes gamblers alike, money is not the main thing. It
is a way of measuring results, a way of keeping score (cf. Alvarez, 1983,
p. 42; McClelland, 1961, p. 237; Thackrey, 1968, pp. 57-58). In the same way,
many currently well-to-do people who grew up in the Great Depression continue
to value security in a way that their more fortunate children cannot
understand. A penny saved is not a penny saved; it is security in the bank.

Nor should custom and duty be forgotten in their effects on risk taking.
Consider the story of the Reindeer and the Montcalm (Mowat, 1982). In March
of 1932 the salvage tug Reindeer set to sea carrying 28 men in a furious storm
to rescue a damaged freighter. Reindeer was not designed for such work and
she was old and ill-equipped. Within hours she was foundering 60 miles from
land. The vessel Montcalm, meanwhile, had been damaged herself by the storm
and was running for harbor. When it became clear that no other ship could
reach Reindeer in time, the master of Montcalm, Captain Rothwell, turned her
back to sea. "It had been no easy decision. Montcalm carried sixty
passengers and a crew of fifty, and their lives were all in Rothwell's care.
The risk to them was real enough, but the death of Reindeer's men was sure
unless that risk was taken" (Mowat, 1982, p. 46). Although Reindeer sank, all

N.-
+ U+
.,+ML-,...- . . . '- .- I:-,.i. ,+. ++ : + . " , ,,. . ,,., -
V.I

Psychology of Risk / 31

were saved, even the ship's dog.

To understand such events requires a more comprehensive view than can be

provided by the simple psychophysics of lives saved or lives lost (cf.
Tversky & Kahneman, 1981) or even by the machinery of hope and fear. Reindeer
was there because risk is the essence of salvage work. Rothwell was captain
of Montcalm because he was capable of exercising the traditional duty that
sailors bear to other sailors. Theories that attempt to explain all of risky
choice in the narrow terms of purely perceptual or purely cognitive or purely
motivational mechanisms will necessarily miss much of what impels people
toward or away from particular risks. The factors that influence human risk
taking range from psychophysics to society and from fear to fun. So too
should the psychology of risk.

i
I°

S"

P' 5
 Psychology of Risk / 32

References

Abramson, L. Y., Seligman, M. P., & Teasdale, J. D. (1978). Learned

helplessness in humans: Critique and reformulation. Journal of Abnormal
Psycholoy, 87, 49-74.
Allais, M. (1979). The foundations of a positive theory of choice involving
risk and a criticism of the postulates and axioms of the American School.
In M. Allais & 0. Hagen (Eds.), Expecte utility hypotheses and the
Allais Paradox (pp. 27-145). Dordrecht, Holland: Reidel. (Original work
published 1952) .
Alvarez, A. (1983). The bigcest ame in town. Boston MA: Houghton Mifflin.
Anderson, J. R. (1979). Perspective on models of uncertain decisions. In In
J. A. Roumasset, J.-M. Boussard, & I. Singh (Eds.), Risk, uncertainty,
and agricultural development (pp. 39-62). New York: Agricultural
Development Council.
Arrow, K. J. (1971). Essays in the the of risk-bearing. Chicago:
Markham.b v
Atkinson, A. B. (1970). On the measurement of inequality. Journal of.-
Economic Theo , .2,244-263.
Atkinson, J. W. (1957). Motivational determinants of risk-taking behavior. -
Psycholoical Review, 64, 359-372.
Atkinson, J. W. (1958). Towards experimental analysis of human motivation in
terms of motives, expectancies, and incentives. In J. W. Atkinson (Ed.),
Motives in fantasy, action, and society (pp. 288-305). Princeton NJ:
Van Nostrand.
Atkinson, J. W. (1983). Personality, motivation, and action. New York:
Praeger.
Bell, D. E (1982). Regret in decision making under uncertainty. Operations
Research, 30, 961-981.
Bell, D. E. (1985). Disappointment in decision making under uncertainty.
Operations Research, 33, 1-27).
Bowman, E. H. (1982). Risk seeking by troubled firms. Sloan Management
Review, 23, 33-42.
Bernoulli, D. (1967). Exposition of a new theory on the measurement of risk.
Farnsborough Hants, England: Gregg Press. (Original work published 1738)
Coombs, C. H. (1975). Portfolio theory and the measurement of risk. In M. F.
Kaplan & S. Schwartz (Eds.), Human udament and decision processes. New
York: Academic Press.
Coombs, C. H., & Avrunin, G. S. (1983). Single-peaked functions and the
theory of preference. Psychological Review, 84, 216-230.
Coombs, C. H., & Lehner, P. E. (1981). Evaluation of two alternative models
of a theory of risk: I. Are moments of distributions useful in
assessing risk? Journal of Experimental Psychology: Human Perception
and Performance, 7, 1110-1123.
Dahlby, B. G. (1985). Ranking income distributions in a Harsanyi framework.
Research paper 85-12, Department of Economics, University of Alberta.
Daston, L. J. (1980). Probabilistic expectation and rationality in classical
probability theory. Historia Mathematica, 7, 234-260.
Day, R. H. (1979). Cautious suboptimizing. In J. A. Roumasset, J.-M.
Boussard, & I. Singh (Eds.), Risk, uncertainty, and agricultural
development (pp. 115-130). New York: Agricultural Development Council.
Douglas, M., & Wildavsky, A. (1982). Risk and culture. Berkeley CA:
University of California.

. . ....
..-..... .., ....- .... . ..•...-.. .-. . . .., . ° .. .. . . . .. . .. . .. ... . .. .. - .. . .. - . .. - .
 I. W."i"
k- 07 2F~ 2 ' ~ ~ .i% 19IVATV

Psychology of Risk / 33

Ellsburg, D. (1961). Risk, ambiguity, and the Savage axioms. Quarterly

Journal of Economics, 75, 643-669.
Ericsson, K. A., & Simon, H. A. (1980). Verbal reports as data.
Psychological Review, 87, 215-251.
Festinger, L. (1957). A theory of coQnitive dissonance. Evanston, IL: Row
Peterson.
Friedman, M. (1953). Essays in positive economics. Chicago: University of
Chicago.
Friedman, M., & Savage, L. J. (1948). The utility analysis of choices
involving risk. Journal of Political Economy, 56, 279-304.
Hagen, 0. (1969). Separation of cardinal utility and specific utility of risk
in theory of choices under uncertainty. Saertrykk av Statsokonomisk
Tidsskrift, 3, 81-107.
Hagen, 0. (1979). Towards a positive theory of preferences under risk. In M.
Allais & 0. Hagen (Eds.), Expect utility hypotheses and the Allais
Paradox (pp. 271-302). Dordrecht, Holland: Reidel.
Hiroto, D. S. (1974). Locus of control and learned helplessness. Journal of
Experimental Psychology, 102, 187-193.
Hiroto, D. S., & Seligman, M. E. P. (1975). Generality of learned
helplessness in man. Journal of Personality and Social Psychology, 31,
311-327.
Janis, I. L., & Mann, L. (1977). Decision making: A psychological analysis
of conflict, choice, and comit. New York: Free Press.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision
under risk. Econometrica, 47, 263-291.
Kahneman, D., & Tversky, A. (1984). Choices, values, and frames. American
Psychologist, 39, 341-350.
Karmarkar, U. S. (1978). Subjectively weighted utility: A descrintive
extension of the expected utility model. Organizational Be&_vior and
Human Performance, 21, 61-82.
Keller, L. R., Sarin, R. K., & Weber, M. (in press). Empirical
investigatioion of some properties of the perceived riskiness of gambles.
Organizational Behavior and Human Decision Processes.
Keyes, R. (1985). Chancing it. Boston MA: Little, Brown.
Kunreuther, H., & Wright, G. (1979). Safety-first, gambling, and the
subsistence farmer. In J. A. Roumasset, J.-M. Boussard, & I. Singh
(Eds.), Risk, uncertainty, and agricultural development (pp. 213-230).
New York: Agricultural Development Council.
Langer, E. J. (1977). The psychology of chance. Journal for the Theory of
Social Behavior, 7, 185-207.
Liberman, A. M., Harris, K. S., Hoffman, H. S., & Griffith, B. C. (1957). The
discrimination of speech sounds within and across phoneme boundaries.
Journal of Experimental Psycholoay, 54, 358-368.
Lopes, L. L. (1981). Decision making in the short run. Journal of
Experimental Psychology: Human Learning and Memor, 7, 377-385.
Lopes, L. L. (1983). Some thoughts on the psychological concept of risk.
Journal of Experimental Psycholoqy: Human Perception and Performance, 9,
137-144.
Lopes, L. L. (1984). Risk and distributional inequality. Journal of
Experimental Psychology: Human Perception and Performance, 10, 465-485.
Lopes, L. L. (1986). Reasoning and risk aversion. Manuscript in preparation.
Lopes, L. L., & Casey, J. (1986). Risk-taking under competition. Manuscript
in preparation.

7,
 Psychology of Risk / 34

Luce, R. D. (1980). Several possible measures of risk. Theory and Decision

12, 217-228.
Machina, M. J. (1982). "Expected utility" analysis without the independence
axiom. Econometrica, 50, 277-323.
Maier, S. F., & Seligman, M. E. P. (1976). Learned helplessness: Theory and
evidence. Journal f Experimental Psychology: General, 105, 3-46.
March, J. G. (1978). Bounded rationality, ambiguity, and the engineering of
choice. Bell Journal of Economics, 9, 587-608.
Markowitz, H. (1952). The utility of wealth. Journal of Political Economy,
60, 151-158.
Markowitz, H. M. (1959). Portfolio selection: Efficient diversification of
investments. New York: Wiley.
Maslow, A. H. (1954). Motivation and personality. New York: Harper and Row.
McClelland, D. C. (1958). Risk-taking in children with high and low need for
achievement. In J. W. Atkinson (Ed.) Motives in fantasy, action, and
society, pp. 306-321. Princeton NJ: Van Nostrand.
McClelland, D. C. (1961). The achieving society. Princeton NJ: Van Nostrand.
Mischel, W. (1968). Personality and Assessment. New York: Wiley.
Mowat, F. (1982). Grey seas under. New York: Bantam.
Ortiz, S. (1979). The effect of risk aversion strategies on subsistence and
cash crop decisions. In J. A. Roumasset, J.-M. Boussard, & I. Singh
(Eds.), Risk, uncertainty, and aQricultural development (pp. 231-246).
New York: Agricultural Development Council.
Pollatsek, A., & Tversky, A. (1970). A theory of risk. Journal of
Mathematical Psychology, 7, 540-553.
Pope, R. (1983). The pre-outcome period and the utility of gambling. In M.
Allais & 0. Hagen (Eds.), Expect utility hypotheses and the Allais
paradox (pp. 138-177). Dordrecht, Holland: Reidel.
Pratt, J. W. (1964). Risk aversion in the small and in the large.
Econometrica, 32, 122-135.
Samuelson, P. A. (1977). St. Petersburg paradoxes: Defanged, dissected, and
historically described. Journal of Economic Literature, 15, 24-55.
Savage, L. J. (1954). The foundations of statistics. New York: Wiley.
Schneider, S. L., & Lopes, L. L. (in press). Reflection in preferences under
risk: Who and when may suggest why. Journal of Experimental Psychology:
Human Perception and Performance.
Siegel, S. (1957). Level of aspiration and decision making. Psychological
Review, 64, 253-262.
Simon, H. A. (1955). A behavioral model of rational choice. Quarterly
Journal of Economics, 69, 99-118.
Slovic, P., Fischhoff, B., & Lichtenstein, S. (1980). Facts and fears:
Understanding perceived risk. In R. C. Schwing & W. A. Albers, Jr.
(Eds.), Societal risk assessment: How safe is safe enou h?. New York:
Plenum.
Thackrey, T. (1968). Gambling secrets of Nick the Greek. Chicago: Rand
McNally.
Thompson, M. (1980). Aesthetics of risk: Culture or context. In R. C.
Schwing & W. A. Albers, Jr. (Eds.), Societal risk assessment: How safe is
safe enough?. New York: Plenum.
Tversky, A., & Kahneman, D. (1981). The framing of decisions and the
psychology of choice. Science, 2:11, 453-458.
von Neumann, J., & Morgenstern, 0. (1947). Theo of game and economic
behavior. Princeton NJ: Princeton University. (2ed edition)

• 7
" 1
 Psychology of Risk / 35

Williams, A. C. (1966). Attitudes toward speculative risks as an indicator of

attitudes toward pure risks. Journal of Risk and Insurance, 33, 577-586.

.9 .~ 9,
7
 Psychology of Risk / 36

Footnotes S
The writing of this article and the research reported in it were
N00014-84-K-0065. I am
grateful to by
facilitated Office
Leonard Berkowitz, Patricia contract
of Naval Research Devine, Mary Douglas, Robin Keller,
David Messick, Gregg Oden, Sandra Schneider, and Alex Wearing for their
helpful criticism and coffnts and (again) to Gregg Oden for his MacWonderful
help with the graphics.

1. Kahneman and Tversky (1979) call their function a "value function" to

distinguish it from the utility function of modern expected utility theory
(von Neumann & Morgenstern, 1947). Because this article deals primarily with
the psychophysical or classical interpretation of utility, the single word
"utility" is used throughout to refer to the subjective value of money or
other commodities.

2. Originally the value was assumed to be .50, but now it is believed to

be nearer .35 (J.Atkinson, 1983).

3. Researchers presumably also have limitations in their ability to

process information. This might be an additional reason for preferring two-
outcome gambles, though it doesn't get mentioned.

4. Hagen's (1969) theory is expressed in terms of the moments of the

distribution of psychological or subjective values. In the present case,
however, the relations among the moments of the objective distributions are
similar to those that would obtain for the subjective distributions.

5. The variance is unchanged when a positive constant is added to each

outcome in the distribution. Risk, however, decreases (Keller, Sarin, &
Weber, in press). For example, if $1,000,000 were added to each of the
outcomes in the various lotteries in Figure 1, their variances would be
unchanged, but they would have become much more similar in terms of perceived
riskiness. Would that we each could participate in such delightfully unrisky
ventures!

6. The graphical analyses that are used in this paper apply only to
lotteries having equal expected value. Although Lorenz curves can be drawn
for lotteries that differ in expected value (by omitting the normalization
step in column 7), comparisons of such Lorenz curves would need to take this
into account. In such cases, comparison by means of a mathematical index of
security or potential would probably be preferable (see Section IV.A).

7. It is worth noting that strategies that involve maximizing the

probability of meeting a goal or aspiration level are fundamentally different
from strategies of maximizing expected utility. In the utility formulation,
necessity can only be captured by assuming that the utility function
temporarily becomes positively accelerated in the region of the target value
(see, e.g., Kahneman & Tversky, 1979, p. 279). Such explanations are
obviously ad hoc since they can be called into play anytime the standard
psychophysics of the situation cannot explain the preference. The alternative
view that subjects sometimes attempt to maximize the probability of achieving
aspiration levels has been taken by Allais (1979) and by Lopes (1981). _
Although this view seems to make intuitive sense, it violates the axioms of
 nu... L - -

Psychology of Risk / 37

expected utility theory and has been considered to be irrational for that
reason (see, e.g., Samuelson, 1977, p. 48).

8. In cases of strong conflict involving extremely important outcomes

(e.g., health issues, large financial transactions, career changes, etc.)
conflict may be reduced by various psychological bolstering processes
(Festinger, 1957; Janis & Mann, 1977). Whether these entail distortion of
values and probabilities during the choice process or selectional mechanisms
operating in the construction of a post-decisional rationalization for the
chosen alternative is an important question, but not one to which the present
experiments can speak. Nevertheless, even if there is considerable distortion
pre-decisionally, security/potential and aspiration level would still function
in people's deliberations about the presumably distorted distributions.

9. One could, of course, explain individual differences within a

psychophysical theory by supposing that people with different preferences have
different utility and probability functions. This would, however, vitiate the
claim that risky choices can be explained by basic perceptual processes. At
the limit, the functions would become a means for surmmarizing preferences
after the fact (as is the case for the von Neumann and Morgenstern, 1947,
utility function), but such functions would lack predictive and explanatory
power.

ix

C.

..
o...
. :<........,
... ..... ...
,....
. .....
.......
. ..< .-¢ . .. . . .........
..
< .. ...
 Psychology of Risk / 38

Table 1
Hypothetical Preferences for Risk Averse and Risk Seeking Individuals
for Gain and Loss Lotteries

Risk Averse Individual

Gain Lotteries Loss Lotteries

LOT AL SEC SEC x AL REL PREF AL SEC SEC x AL REL PREF

ST 1.00 1.00 1.000 .310 0.00 1.00 .000 .000

RL 1.00 0.83 0.830 .257 0.00 0.58 .000 .000
SS 0.90 0.58 0.522 .162 0.10 0.83 .083 .284
PK 0.90 0.58 0.522 .162 0.10 0.58 .058 .199
RC 0.75 0.33 0.248 .077 0.25 0.33 .083 .284
BM 0.60 0.17 0.102 .032 0.40 0.17 .068 .233
LS 0.47 0.00 0.000 .000 0.53 0.00 .000 .000

Risk Seeking Individual

Gain Lotteries Loss Lotteries

LOT AL POT POT x AL REL PREF AL POT POT x AL REL PREF

ST 1.00 0.00 0.000 .000 0.00 0.00 .000 .000

RL 0.69 0.42 0.290 .143 0.00 0.17 .000 .000
SS 0.78 0.17 0.133 .065 0.03 0.42 .013 .022
PK 0.74 0.42 0.311 .153 0.02 0.42 .008 .013
RC 0.60 0.66 0.396 .195 0.10 0.66 .066 .111
BM 0.52 0.83 0.432 .213 0.24 0.83 .199 .333
LS 0.47 1.00 0.470 .231 0.31 1.00 .310 .520

Note. Rows are ST (sure thing), RL (riskless lottery), SS (short shot), PK

(peaked lottery), RC (rectangular lottery), BM (bimodal lottery), and LS
(long shot). AL is probability of achieving the aspiration level. SEC is
security and POT is potential. REL PREF is relative preference.

~ * ~ ~ b *
 Psychology of Risk / 39

Figure Captions

1. Examples of utility functions with four different shapes. The

Bernoullian function (upper left) is uniformly risk averse (negatively
accelerated). The functions in the upper right, lower left, and lower right
(suggested by Friedman and Savage, 1948, Markowitz, 1952, and Kahneman and
Tversky, 1979, respectively) have regions of risk aversion (negative
acceleration) and risk seeking (positive acceleration). The upper two
functions range from zero assets to large positive assets. The lower two
functions range about a customary asset level (e.g., the status quo).

2. Examples of stimulus lotteries for gains. Each lottery has 100

tickets (represented by tally marks) and each has an expected value of
approximately $100. The values at the left give the prizes that are won by
tickets in that row. The lotteries are ordered from the upper left to the
lower right in the order that they are preferred by risk averse subjects.

3. Demonstration of how to draw the Lorenz curve for a lottery. The

panel on the left shows how probabilities and relative gains can be cumulated
for the peaked lottery. The panel on the right gives the Lorenz curve. This
plots the cumulative probability on the abscissa and the cumulative proportion
of the winnings on the ordinate.
4. Comparison of the Lorenz curves of the long shot (LS) and the short
shot. Lorenz curves that lie near the diagonal at the low end (stippled area)
have relatively few small outcomes. Lorenz curves that lie far from the
diagonal at the high end (striped area) have a few extremely large outcomes or
relatively many moderately large outcomes.

5. Comparison of the Lorenz curves for the riskless lottery (RL) and the
short shot (SS). The fact that the curves cross one another indicates that RL
is good both for avoiding small outcomes (stippled area) and for approaching
large outcomes (striped area).

6. Comparison of Lorenz curves for the bimodal lottery (BM) and the long
shot (LS). These lotteries are very similar at their low ends (stippled area)
but very different at their high ends (striped area).

7. Comparison of Lorenz curves for the bimodal (BM) and long shot (LS)
loss lotteries. These lotteries are very different at their low ends
(stippled area) but very similar at their high ends (striped area).

8. Comparison of Lorenz curves for the riskless (RL) and the short shot
(SS) loss lotteries. The fact that the curves cross one another indicates
that SS is good both for avoiding large losses (stippled area) and for
approaching small or zero losses (striped area).

9. Mean preference data for risk averse subjects for gain lotteries
(open symbols) and loss lotteries (filled symbols). Data are the number of
times a subject chose a lottery out of the total number of times the lottery
was available for choice. (From "Reflection in preferences under risk: Who
and when may suggest why" by S. L.& Schneider and L. L. Lopes, Journal of
Experimental Psycholoc1 : Human Perception and Performance, in press.
Copyright 1986 by The American Psychological Association. Reprinted by
 Psychology of Risk / 40

permission.)
10. Mean preference data for risk seeking subjects for gain lotteries
(open symbols) and loss lotteries (filled symbols). Data are the number of
tines a subject chose a lottery out of the total number of times the lottery
was available for choice. (From "Reflection in preferences under risk: Who
and when may suggest why" by S. L.& Schneider and L. L. Lopes, Journal of
Experimental Psycholoiy: Human Perception and Performance, in press.
Copyright 1986 by The American Psychological Association. Reprinted by
permission.)

.I.o

I.

I.

.4.
 DOLLARS
DOLLARS

customary
position'**

LOSSES GAINS
DOLLARS

,°

... ... ,, ... , -,.-.. , .. . ..,... 4.

..,..
.... .. .,... .,..,.-..
.,- ....-,. ...
.. ., ... -. . ..
, ,.,

Figure 1[
 $200 I
$186 I
$172 III
$159 11111
$146 1111111
$200 I $130 IIIIIIIIIIIIIIIIIIIIIIIIIIIIIII $132
$119 111111111
IlIlIIIII
$187 II $115 IIIIIIIIIIIIIIIIIIIIII $106 I IIII
$172 I1! $101 IIIIIIIIIIIII $ 93 IIIII
$157 Ilii $ 86 IlIIIII $ 80 IlIIIIII
$143 1IIII $ 71 1111111 $ 66 Ill9ll3ll
$129 IIIIIII $ 57 11111 $ 53 1111111
$114 $ 43
$IIIIIIIII 5III 11111
$ 40
$ 99 IIIIIIIIIIIIIII $ 28 III $26 III
$ 85 IIIIIIIIIIIIIIIIIIIII $ 13 II $13
$ 70 IIIIIIIIIIIIIIIIIIIIIIIIIIIIII ZERO I ZERO I

RISKLESS SHORT SHOT PEAKED

$200 11111
$189 i11
$178 JIIII
$168 11111
$158 11111 $200 IIIIIIIIII
$147 11111 $186 IIIIIIII
$136 11111 $172 111111111
$126 11111 $159 1111111
$116 11111 $146 11111
$105 11111 $132 Ill
$ 94 11111 $119 I $439 I
$ 84 11111 $106 I $390 II
$ 74 11111 $ 93 I $341 III
$ 63 1IIII $ 80 I $292 1I11
$ 52 11111 $ 66 III $244 11111
S 42 11111 S 53 11111 $195 1111111
$ 32 11111 $ 40 1111111 $146 II
$ 21 11111 $ 26 111111111 $ 98 IIIIIIIIIIIIII
$ 10 11111 $ 13 IIIIIIIIS 49 IIIIIIIIIIIII
ZERO 11111 ZERO IIIIII ZERO IIIIIIIIIIIIIIIIIIIIIIIIIIIIi '
RECTANGULAR BIMODAL LONG SHOT

Figure 2

:.:: . .- ..-.
.-. .- ."--, ". .."- ."- v -." •" -. '- -. - .--. . -".
-" ,' "- -, " ". - - -" "- . -" "
 CUMULATIVE VALUE

cc:J

x x
a_ Q-w Lnr- L N W M M C> )

C)0 ci) - cv( r 0Ci li( (0 0 0

C:
m qRT : D I-c M
a MC 0-
D . . . . . . . . . ..

~~O~O~viO
oo>
)C~-oc'(DLnC)'Joa q como

0Q .0 . . . . . . . . .

a.

Cd)

LJ
0 0 CLeJ'o
RT w--aMq ---- n

V'. ~ ~ *I*U ).*.* ~'.* . . AA.... * *'a_-

1%*
7677

.14.

.........

..
.Q.
. . ..

.. .. . . . . . . .
 0

. .. .. . . . .

*. . . . .. .~ . .. . .. .

.5** 5*...........5

... ~ ~XI .. .. . .e .. ... ........

... .... .. ...... ....

.5* /* at........./.

.. . . .. . .
./.. .. ..
 CUMULATIVE VALUE

/...............
... S..........
.

... . . . .

......
...
.... 0.

~ ~ K.....................
 .
. / ./

.... .... .... .... ..

./~~... *////* .....
... ... ..
.. . .. .
.. . .........

/.'/* //o/

.. . . .....
..... .. . . .. . .
-.

0 ~~ flVA ALLv'1hNflo

~ ... *.- **~*......................

 wtt

de,

do P. op / /

" /cr
//0

............
.........
A. . .......
... ......... 5
........
.
. .. .......-

........ w ..
.. ....
......
...

......
...
D..
.
...
D.....
0..... ...........

0 BflVABAiivflV\JfO. ...
....
....
....
....

1p 4
~~~~V~~F
... Vw 1 r I1WVI LIU L.W 7V. UU-VV"W rw %- yw- Im VYV ~ 1'y~

coo

Clz)

w 0
Cof

CQ

'4 soq sUum joI-.xN vp

 0 C0

CO'
c-J

C'))

CO

o to~ 0 itO o io a LO 0 In~ 0

16 C~i 6 r 1- C~i 6 r. 6 C~i

tiasoiqo .etuq ;o .zaqumm ua?(

 January 1986
.,

OFFICE OF NAVAL RESEARCH oo

Engineering Psychology Program

TECHNICAL REPORTS DISTRIBUTION LIST

OSD
CAPT Paul R. Chatelier J. Randy Simpson
Office of the Deputy Under Secretary Statistics Program Code 1111SP
of Defense Office of Naval Research
OUSDRE (E&LS) 800 North Quincy Street
Pentagon, Room 3D129 Arlington, VA 22217-5000
Washington, D. C. 20301 Dr. Lyle 0. Broemeling
Department of the Navy Code 1111SP
Engineering Psychology Program Office of Naval Research
Office of Naval Research 800 N. Quincy Street
Code 1142EP Arlington, VA 22217-5000
800 North Quincy Street
Arlington, VA 22217-5000 (3 copies) Information Sciences Division
Code 1133
Aviation & Aerospace Technology Office of Naval Research
Programs 800 North Quincy Street
Code 121 Arlington, VA 22217-5000
Office of Naval Research
800 North Quincy Street CAPT William M. Houk
Arlington, VA 22217-5000 Commanding Officer
Naval Medical R&D Command
Physiology and Neurobiology Program Bethesda, MD 20814-5055
Office of Naval Research
Code 1141NP Dr. Randall P. Schumaker
800 North Quincy Street NRL A. I. Center
Arlington, VA 22217-5000 Code 7510
Naval Research Laboratory
CAPT. P. M. Curran Washington, D.C. 20375-5000
Code 125
Office of Naval Research
800 North Quincy Street
Arlington, VA 22217-5000

Dr. Charles Holland

Office of Naval Research
Code 1133
800 N. Quincy Street
Arlington, VA 22217-5000
 JANUARY 1986

Department of the Navy

Special Assistant for Marine Dr. James McMichael

Corps Matters Office of the Chief of Naval
Code OOMC Operations, oP987H
Office of Naval Research Technology Assessment Division
800 North Quincy Street Washington, D. C. 20350
Arlington, VA 22217-5000 M Jna
Mr. John Davis'"
Mr. R. Lawson Combat Control Systems Department
ONR Detachment Code 35
1030 East Green Street Naval Underwater Systems Center
Pasadena, CA 91106-2485 Newport, RI 02840

CDR James Offutt Human Factors Department

Office of the Secretary of Defense Code N-71
Strategic Defense Initiative Organization Naval Training Systems Center
Washington, D.C. 20301-7100 Orlando, FL 32813
Director Mr. Norm Beck
Technical Information Division Combat Control Systems Department
Code 2627 Code 35
Naval Research Laboratory Naval Underwater Systems Center
Washington, D.C. 20375-5000 Newport, RI 02840

Dr. Michael Melich Human Factors Engineering

Communications Sciences Division Code 441
Code 7500 Naval Ocean Systems Center
Naval Research Laboratory San Diego, CA 92152
Washington, D.C. 23075-5000
Dr. Gary Poock
Operations Research Department
Dr. J. S. Lawson, Jr. Naval Postgraduate School
4773-C Kahala Avenue Monterey, CA 93940
Honolulu, HI 96816
Mr. H. Talkington
Engineering & Computer Science
Dr. Neil McAlister Code 09
Office of Chief of Naval Operations Naval Ocean Systems Center
Command and Control San Diego, CA 92152
OP--O94H
Washington, D. C. 20350 CDR Paul Girard
Command & Control Technology
Dr. A. F. Norcio Department, Code 40
Computer Sciences & Systems Naval Ocean Systems Center
Code 7592 San Diego, CA 92152
Naval Research Laboratory
Washington, DC 20375-5000

2
 JANUARY 1986

Department of the Navy

Mr. Paul Heckman Commander

Naval Ocean Systems Center Naval Air Systems Command
San Diego, CA 92152 Crew Station Design
NAVAIR 5313
Dr. William Uttal Washington, D. C. 20361
Naval Ocean Systems Center
Hawaii Laboratory Mr. Philip Andrews
P. 0. Box 997 Naval Sea Systems Command
Kailua, HI 96734 NAVSEA 61R
Washington, D. C. 20362
Dr. A. L. Slafkosky
Scientific Advisor Aircrew Systems Branch
Commandant of the Marine Corps Systems Engineering Test
Washington, D. C. 20380 Directorate
U.S. Naval Test Center
Dr. L. Chmura Patuxent River, MD 20670
Computer Sciences & Systems
Code 7592 Mr. Milton Essoglou
Naval Research Laboratory Naval Facilities Engineering
Washington, D.C. 20375-5000 Command
R&D Plans and Programs
Dr. Michael Letsky Code 03T
Office of the Chief of Naval Hoffman Building II
Operations (OP-O1B7) Alexandria, VA 22332
Washington, D.C. 20350
CAPT Robert Biersner
Professor Douglas E. Hunter Naval Biodynamics Laboratory
Defense Intelligence College Michoud Station
Washington, D.C. 20374 Box 29407
New Orleans, LA 70189
CDR C. Hutchins
Code 55 Dr. Arthur Bachrach
Naval Postgraduate School Behavioral Sciences Department
Monterey, CA 93940 Naval Medical Research Institute
Bethesda, MD
Dr. Stanley Collyer
Office of Naval Technology Dr. George Moeller
Code 222 Human Factors Engineering Branch
800 North Quincy Street Naval Submarine Base
Arlington, VA 22217-5000 Submarine Medical Research Lab.
Groton, CT 06340
Professor Michael Sovereign
Joint Command, Control &
Communications Curriculum
Code 74
Naval Postgraduate School
Monterey, CA 93943

3
 JANUARY 1986

Department of the Navy

Head Dean of the Academic Departments

Aerospace Psychology Department U. S. Naval Academy
Naval Aerospace Medical Research Lab Annapolis, MD 21402
Pensacola, FL 32508
CDR W. Moroney
Naval Air Development Center
Commanding Officer Code 602
Naval Health Research Center Warminster, PA 18974
San Diego, CA 92152
Dr. Harry Crisp
Dr. Jerry Tobias Code N 51
Auditory Research Branch Combat Systems Department
Submarine Medical Research Lab Naval Surface Weapons Center
Naval Submarine Base Dahlgren, VA 22448
Groton, CT 06340
Mr. John Quirk
Dr. Robert Blanchard Naval Coastal Systems Laboratory
Code 71 Code 712
Navy Personnel Research and Panama City, FL 32401
Development Center
San Diego, CA 92152-6800 Human Factors Branch
Code 3152
LCDR T. Singer Naval Weapons Center
Human Factors Engineering Division China Lake, CA 93555
Naval Air Development Center
Warminster, PA 18974 CDR Kent S. Hull
MS 239-21
Mr. Jeff Grossman NASA/Ames Research Center
Human Factors Division, Code 71 Moffett Field, CA 94035
Navy Personnel R&D Center
San Diego, CA 92152-6800 De Rabinder N. Madan
Dr. -.
LT. Dennis McBride Code 1114SE
Human Factors Branch Office of Naval Research
Pacific Missle Test Center 800 North Quincy Street
Point Mugu, CA 93042 Arlington, VA 22217-5000 .

Dr. Kenneth L. Davis Dr. Eugene E. Gloye

Code 1114 ONR Detachment
Office of Naval Research 1030 East Green Street
800 North Quincy Street Pasadena, CA 91106-2485
Arlington, VA 22217-5000
LCDR R. Carter
Office of Chief on Naval Operations
(OP- 1B)
Washington, D.C. 20350

r %

4
v w
 JANUARY 1986

Dr. Glen Allgaier Department of the Air Force

Artificial Intelligence Branch

Code 444 Dr. Kenneth R. Boff
Naval Electronics Ocean System Center AF AMRL/HE
San Diego, CA 921525 Wright-Patterson AFB, OH 45433

Dr. Steve Sacks Dr. A. Fregly

Naval Electronics Systems Command U.S. Air Force Office of
Code 61R Scientific Research
Washington, D.C. 20363-5100 Life Science Directorate, NL
Bolling Air Force Base
Dr. Sherman Gee Washington, D.C. 20332-6448
Command and Control Technology, (Code 221)
Office of Naval Technology, Mr. Charles Bates, Director a,
800 N. Quincy Street Human Engineering Division
Arlington, VA 22217-5000 USAF AMRL/HES
Wright-Patterson AFB, OH 45433
Dr. Robert A. Fleming
Human Factors Support Group Dr. Earl Alluisi p.

Naval Personnel Research & Development Ctr. Chief Scientist

1411 South Fern Street AFHRL/CCN
Arlington, VA 22202 Brooks Air Force Base, TX 78235

Department of the Army Dr. J. Tangney

Directorate Life Sciences
Dr. Edgar M. Johnson AFSOR
Technical Director Bolling AFB
U.S. Army Research Institute Washington, D.C. 20032-6448
Alexandria, VA 22333-5600
Mr. Yale Smith
Technical Director Rome Air Development
U.S. Army Human Engineering Laboratory Center, RADC/COAD
Aberdeen Proving Ground, MD 21005 Griffiss AFB
New York 13441-5700
Director, Organizations and Systems
Research Laboratory Dr. A. D. Baddeley
U.S. Army Research Institute Director, Applied Psychology
5001 Eisenhower Avenue Unit
Alexandria, VA 22333-5600 Medical Research Council
15 Chaucer Road
Dr. Milton S. Katz Cambridge, CB2 2EF England
Director, Basic Research
Army Research Institute Dr. Kenneth Gardner
5001 Eisenhower Avenue Applied Psychology Unit
Alexandria, VA 22333-5600 Admiralty Marine Tech. Estab.
Teddington, Middlesex
TW11 OLN
England

p
" S
 JANUARY 1986

Other Government Agencies

Dr. M. C. Montemerlo Dr. Donald D. Hoffman

Information Sciences & University of California
Human Factors Code RC (Irvine)
NASA HQS School of Social Sciences
Washington, D.C. 20546 Irvine, CA 92717

Dr. Alan Leshner Dr. T. B. Sheridan

Deputy Division Director Dept. of Mechanical
Division of Behavioral and Engineering
Neural Sciences Massachusetts Institute of
National Science Foundation Technology
1800 G. Street, N.W. Cambridge, MA 02139
Washington, D.C. 20550 Dr. Daniel Kahneman
Defense Technical Information The University of British
Center Department of Psychology
Cameron Station, Bldg. 5 #154-2053 Main Mall
Alexandria, VA 22314 (2 copies) Vancouver, British Columbia
Canada V6T 1Y7
Dr. Clinton Kelly
Defense Advanced Research Dr. Stanley Deutsch
Projects Agency NAS-National Research Council
1400 Wilson Blvd. (COHF)
Arlington, VA 22209 2101 Constitution Avenue, N.W.
Washington, D.C. 20418

Other Organizations Dr. Meredith P. Crawford

Dr. arry Snyder American Psychological
Dept. of Industrial Engineering Association
Virginia Polytechnic Institute Office of Educational Affairs
and State University 1200 17th Street N.W.
Blacksburg, VA 24061 Washington, D.C. 20036

Dr. Amos Tversky Dr. Deborah Boehm-Davis

Dept. of Psychology Department of Psychology
Stanford University George Mason University
Stanford, CA 94305 4400 University Drive
Fairfax, VA 22030
Dr. Amos Freedy
Perceptronics, Inc.
6271 Variel Avenue
Woodland Hills, CA 91364

Dr. Jesse Orlansky

Institute for Defense Analyses
1801 N. Beauregard Street
Alexandria, VA 22311

6 .- .
 JANUARY 1986

Other Organizations

Dr. David Van Essen Dr. Stanley N. Roscoe

California Institute of Tech. New Mexico State University
Division of Biology Box 5095
Pasadena, CA 91125 Las Cruces, NM 88003

Dr. James H. Howard, Jr. Mr. Joseph G. Wohl

Department of Psychology Alphatech, Inc.
Catholic University 3 New England Executive Park
Washington, D.C. 20064 Burlington, MA 10803
Dr. William Howell . Dr. Marvin Cohen
Department of Psychology Decision Science Consortium, Inc.
Rice University Suite 721
Houston, TX 77001 7700 Leesburg Pike
Falls Church, VA 22043
Dr. Christopher Wickens
Department of Psychology Dr. Scott Robertson
University of Illinois Catholic University
Urbana, IL 61801 Department of Psychology
Washington, D.C. 20064
Dr. Robert Wherry
Analytics, Inc. Dr. William B. Rouse
2500 Maryland Road School of Industrial and Systems
Willow Grove, PA 19090 Engineering
Georgia Institute of Technology
Dr. Edward R. Jones Atlanta, GA 30332
Chief, Human Factors Engineering
McDonnell-Douglas Astronautics Co. Ms. Denise Benel
St. Louis Division Essex Corporation
Box 516 333 N. Fairfax Street
St. Louis, MO 63166 Alexandria, VA 22314

Dr. Lola L. Lopes Dr. Andrew P. Sage

Department of Psychology Assoc. V. P. for Academic Affairs
University of Wisconsin George Mason University
Madison, WI 53706 4400 University Drive
Fairfax, VA 22030
Dr. Joaquin Fuster
University of California at Dr. James Ballas
Los Angeles Georgetown University
760 Westwood Plaza Department of Psychology
Los Angeles, CA 90024 Washington, D.C. 20057

7
 JANUARY 1986

Other Organizations

Dr. Richard Pew Dr. Robert A. Hummel

Bolt Beranek & Newman, Inc. New York University
50 Moulton Street Courant Inst. of Mathematical
Cambridge, MA 02238 Sciences
251 Mercer Street
Dr. Hillel Einhorn New York, New York 10012
Graduate School of Business
University of Chicago Dr. H. McI. Parsons
1101 E. 58th Street Essex Corporation
Chicago, IL 60637 333 N. Fairfax Street
Alexandria, VA 22314
Dr. Douglas Towne
University of Southern California Dr. Paul Slovic
Behavioral Technology Lab Decision Research
1845 South Elena Avenue, Fourth Floor 1201 Oak Street
Redondo Beach, CA 90277 Eugene, OR 97401
Dr. James T. Todd Dr. Kent A. Stevens
Brandeis University University of Oregon
Waltham, MA 02254 Dept. of Computer & Info Sci.
Eugene, OR 97403
Dr. John Payne
Graduate School of Business Dr. Donald A. Glaser
Administration U. of California, Berkeley
Duke University Department of Molecular Biology
Durham, NC 27706 Berkeley, CA 94720

Dr. Dana Yoerger

Deep Submergence Laboratory
Woods Hole Oceanographic
Institution
Woods Hole, MA 02543 .

Dr. Azad Madni

Perceptronics, Inc.
6271 Variel Avenue
Woodland Hills, CA 91364
Dr. Tomaso Poggio
Massachusetts Institute of Tech.
Center for Biological Information
Processing
Cambridge, MA 02139
Dr. Whitman Richards
Massachusettes Ins. of Tech
Department of Psychology
Cambridge, MA 02139

8
. . . . . . 9 . . .. . . . . ... .. . .. - . ... -p
 JANUARY 1986

Other Organizations

Dr. Leonard Adelman Dr. Alexander Levis

PAR Technology Corp. Massachusetts Institute of
Building A Technology
1220 Sunset Hills Road, Suite 310 Lab Information & Decision Systems
McLean, VA 22090 Cambridge, MA 02139

Dr. Michael Athans Dr. D. McGregor

Massachusetts Inst. of Technology Perceptronics Inc.
Lab Information & Decision Systems 1201 Oak Street
Cambridge, MA 02139 Eugene, OR 97401

Dr. David Castanon Dr. David Noble

ALPHATECH, Inc. Engineering Research Assoc.
I11 Middlesex Turnpike 8616 Westwood Center Dr.
Burlington, MA 01803 McLean, VA 22180

Dr. A. Ephremides Dr. P. Papantoni-Kazakos

University of Maryland University of Connecticut
Electrical Engineering Dept. Department of Electrical Engin.
College Park, MD 20742 and Computer Science (U-157)
Storrs, CT 06268
Dr. Baruch Fischhoff
Perceptronics, Inc. Professor Wayne F. Stark
6271 Variel Ave. University of Michigan
Woodland Hills, CA 91367 Department of Electrical Eng.
and Computer Science
Dr. Bruce Hamill Ann Arbor, MI 48109
The Johns Hopkins Univ.
Applied Physics Lab Mr. Robert L. Stewart
Laurel, MD 20707 The Johns Hopkins University
Applied Physics Laboratory
Barry Hughes Laurel, MD 20707
Space and Naval Warfare Systems 4

Code 611 Dr. Kepi Wu

Washington, D.C. 20363-5100 Space and Naval Warfare Systems
Code 611
Dr. E. Douglas Jensen Washington, D.C. 20363-5100
Carnegie-Mellon University
Computer Science Dept.
Pittsburgh, PA 15213

Dr. David L. Kleinman

Electrical Engineering &
Computer Science Dept.
University of Connecticut
Storrs, CT 06268

: 9
p _______________________