The Journal of Psychology and Financial Markets Copyright © 2001 by
2001, Vol. 2, No. 3, 160 172 The Institute of Psychology and Markets
Psychological Study of Human Judgment:
Implications for Investment Decision Making
Paul Slovic
Reprinted from THE JOURNAL OF FINANCE, Vol. XXVII, No. 4, September 1972
 You are face it a bunch of emotions, preju- given to the problems of interpreting this information
dices, and twitches, and this is all very well as long as skillfully. Graham et al., in their classic treatise on se-
you know it. Successful speculators do not necessarily curity analysis, recognized the proper use of informa-
have a complete portrait of themselves, warts and all, tion as a key element of investment decision making.
in their own minds, but they do have the ability to stop They observed,  After the analyst has learned what in-
abruptly when their own intuition and what is happen- formation he can get and where to get it, he faces the
ing Out There are suddenly out of kilter. harder question: What use to make of it (Graham,
 If you don t know who you are, this is an expensive Dodd, Cottle and Tatham, 1962; p.85)?
place to find out. Many aspects of investment analysis are said to be
 Adam Smith, The Money Game psychological in nature; certainly, the appraisal of
man s capabilities for integrating information into a
judgment or decision is one such aspect. Because of a
Introduction lack of relevant psychological knowledge, security an-
alysts have all too often been forced to become amateur
Just as the stock market has been described as psychologists themselves. For example, G. A. Drew
 The Money Game, security analysis, whether by asserted in 1941 that,
expert or novice, might aptly be labeled  The In-
formation Game. In no other realm are such vast
 In fact, simplicity or singleness of approach is a
quantities of information from such diverse sources
greatly underrated factor of market success. As soon
brought to bear on so many important decisions.
as the attempt is made to watch a multiplicity of fac-
Careful accumulation and skilled interpretation of
tors, even though each has some element to justify it,
this information is said to be the sine qua non of ac- one is only too likely to become lost in a maze of con-
tradictory implications. & The various factors in-
curate evaluation of securities.
volved may be so conflicting that the conclusion fi-
The basic tenet of those in charge of helping the in-
nally drawn is no better than a snap judgment would
vestor to make market decisions seems to be  the more
have been (Drew, 1941; p. 86).
information, the better. Bernhard, writing in 1959,
noted that,
Is Drew s speculation correct? What are man s
limitations as a processor of information? The pur-
 Large brokerage houses undertook big advertising
campaigns to acquaint investors with the  research pose of this paper is to acquaint the reader with psy-
services. Of what did the research consist? Primarily,
chology s recent endeavors to answer this general
it was represented as a careful compilation of all the
question. Along the way we shall touch on a number
facts deemed relevant to an understanding of the sub-
of related topics, including studies of the accuracy
ject company and its stock. That done, the customer
and reliability of judgment; techniques for modeling
was left to his own devices to evaluate the facts (Bern-
the judgment process and making intuition explicit;
hard, 1959; p. 34).
biases in judgments of probability, variability, and
correlation; experimental studies of risk-taking be-
Modern technology has contributed its share to the
havior; and discussion of the relative merits of scien-
information explosion by making vastly greater quan-
tific versus intuitive approaches to making judgments
tities of elegant data readily available to the analyst,
and decisions. Wherever possible, implications of this
broker, and investor. However, little attention has been
work for investment decision making will be noted.
If, as we proceed, we expose some warts, prejudices,
Paul Slovic is president of Decision Research in Eugene, Oregon.
and twitches, it is done in the belief that a full under-
He is also a professor of psychology at the University of Oregon.
standing of human limitations will ultimately benefit
Requests for reprints should be sent to: Paul Slovic, Decision
the decision maker more than will naïve faith in the
Research, 1201 Oak Street, Eugene, OR 97401. Email: pslovic@
oregon.uoregon.edu infallibility of his intellect.
160
PSYCHOLOGICAL STUDY OF HUMAN JUDGMENT
Scientific vs. Subjective Prediction Judgmental Accuracy
in Finance in Investment Analysis
Gray [1966] has recently warned security analysts
To set the stage for a discussion of the relevant psy-
that unless they develop procedures for measuring the
chological literature, I would like to review briefly cur-
validity of their efforts they are likely to have such as-
rent opinion within the domain of investment analysis
sessments imposed upon them by those outside the
pertaining to the analyst s use of information.
profession. Despite the need for such appraisal, there
The analyst is called upon to make predictions,
have been relatively few attempts to assess the results
forecasts, diagnoses, evaluations, etc., on the basis of
of decisions made by analysts or investors under the
fallible information, and with regard to such qualities
harsh light of scientific scrutiny. And, in these investi-
as expected returns, growth rates, variability, and cor-
gations, the performance of man s inferences has ap-
relation. There is a branch of applied mathematics,
peared rather mediocre.
namely statistics, whose purpose is to help men make
Cowles [1933] made one of the first and most exten-
these kinds of judgments. Most of the time, however,
sive attempts to determine the validity of  expert fore-
we bypass formal statistical procedures when making
casting. He found that sixteen financial services, mak-
judgments, and when we do this we are acting as  intu-
ing some 7500 recommendations on individual stocks
itive statisticians.
between 1928 and 1932, compiled an average record
The relative merit of scientific or statistical vs. sub-
that was worse than that of the average common stock
jective or intuitive methods of prediction is a contro-
by 1.4% annually.
versial issue. The intuitive approach is the traditional
Cowles close analysis of the forecasts of William
and predominant method. Here decisions are seen as
D. Hamilton, editor of the Wall Street Journal, over the
based more or less on a state of mind, on feelings or
26 years between 1904 and 1929, showed that they
attitudes, on knowing, without the conscious use of
achieved a result poorer than a representative sample of
well-defined reasoning. For example, consider the fol-
stocks. Similarly, poor results were achieved by 24 fi-
lowing quotations:
nancial publications between the years 1928 and 1932.
A follow-up study in 1944 produced further negative
 & this is no science. It is an art. Now we have com-
findings along with the observation that more than
puters and all sorts of statistics, but the market is still
80% of all forecasts were bullish despite the fact that
the same and understanding the market is still no eas-
bear markets predominated during the period studied
ier. It is personal intuition, sensing patterns of behav-
(Cowles, 1944).
ior & (Smith, 1968; p. 20).
Treynor and Mazuy [1966] evaluated the perfor-
 What is it the good [money] managers have? It s a
mance of 47 mutual funds and the sensitivity of their
kind of locked-in concentration, an intuition, a feel,
nothing that can be schooled (Treynor and Mazuy, portfolio managers to market fluctuations. They rea-
1966; pp. 25 26).
soned that, if fund managers were able to anticipate
major turns in the stock market, they would adjust the
proportion of high and low-risk securities in their port-
In the opposite corner are advocates of a scientific
folios accordingly. Treynor and Mazuy found no evi-
approach to investment analysis. Bauman [1967] de-
dence of such adjustments and they concluded that per-
fines the scientific approach as one which consistently
haps no investor, professional or amateur, can outguess
applies investment theory or a set of decision rules to a
the market. Several other investigations also indicate
variety of investment situations, taking advantage of
that the market has performed as well or better than a
theoretically-derived or empirically-determined quan-
considerable number of professionally managed funds
titative relationships between market factors and secu-
(Bauman, 1965; Fama, 1965; Jensen, 1968; O Brien,
rity performance. Although subjectivists criticize the
1970).
scientific approaches as being too static and insensitive
Perhaps the most extensive studies of the validity of
to subtle factors, scientific methods are rapidly gaining
individual investors judgments are the several  Value
in popularity due to the availability of sophisticated
mathematical and statistical techniques and the devel- Line Contests in which individuals pit their own port-
opment of high-speed computers by which to imple- folios against those selected by Value Line and against
the market averages. The 1969 contest attracted 65,000
ment them. Lorie [1966] observed that a tremendous
amount of research is in progress on such diverse sub- entrants (Murphy, 1969). There appeared to be a great-
jects as insider trading, the effect of stock splits, port- er number of superior portfolios among these than
folio selection, prediction of stock prices and earnings, could be expected by chance. However, the majority of
etc. He concluded that much of this research work has submitted portfolios were not analyzed in enough de-
already had the effect of discrediting beliefs and even tail to permit a careful evaluation of contestants abili-
some relatively sophisticated ones about the behav- ties. Unfortunately, the results of two earlier Value
ior of security prices. Line Contests were also inconclusive regarding inves-
161
SLOVIC
tors ability to outperform the market averages (Bern- the amount of information available to the judge was not
hard, 1967 and Hausman, 1969). necessarily related to the accuracy of his inferences.
A study by Cragg and Malkiel [1968] examined Typical of these is a study by Oskamp [1965] who had
earnings projections for 185 corporations made by five 32 judges, including eight experienced clinical psy-
different forecasting firms. The correlations between chologists, read background information about a pa-
predicted and actual earnings turned out to be quite tient s case. The information was divided into four sec-
low, leading Cragg and Malkiel to conclude that the tions. After reading each section of the case, the judge
careful, painstaking efforts of the analysts performed answered 25 questions about the personality of the sub-
little better than simple projections of past growth ject. The correct answers were known to the investi-
rates. gator. The clinician also rated his confidence in his
On the basis of this brief review, several conclusions answers. Oskamp found that, as the amount of informa-
seem warranted. First, there have still been relatively tion about the case increased, accuracy remained at
few studies concerned with the forecasting ability of about the same level while the clinicians confidence in-
sophisticated investors and analysts. Second, even creased dramatically and became disproportionately
these studies have not directly addressed the subjective great. Thesefindingsmayexplain,inpart,theprevailing
vs. scientific issue. For example, the extent to which tendency to provide the investment decision maker with
the analysts and fund managers evaluated by Cragg as much information as possible. It makes him feel more
and Malkiel, Cowles, or Treynor and Mazuy used sta- confident, but will it improve his decision? This type of
tistical as opposed to intuitive methods is not known. study would seem worth replicating within the context
With the exception of Cowles work, interest in these of security analysis.
issues is of relatively recent origin. In contrast, these The lack of validity of clinical judgments has led to
and related questions have been studied extensively for a number of studies of their reliability. Goldberg distin-
several decades within psychology and medicine. The guished among three types of reliability: (a) consis-
following discussions of relevant psychological re- tency, or stability across time for the same judge using
search may place the results from securities analysis the same data; (b) consensus, or agreement across
into broader perspective while suggesting worthwhile judges using the same data: and (c) convergence, or
avenues for further investigation. agreement when one judge makes several judgments of
the same case but uses different data each time. Studies
reviewed by Goldberg indicate that consistency tends
Psychological Studies to be moderately high but consensus and convergence
of Human Judgment leave much to be desired. The reliability of investment
forecasts would seem to merit systematic study. For
Psychologists have a dual reason for studying the one such attempt see (Cragg and Malkiel, 1968).
types of judgmental processes involved in stock market Even in medicine, studies of clinical judgment have
decisions. First, such processes are obviously within often revealed a surprising degree of unreliability and
the proper domain of psychological inquiry. Second, inaccuracy. Bakwin [1945] reported an experiment
clinical psychologists are forced to use similar pro- which showed that there was no correlation whatsoever
cesses in their role as diagnosticians of mental disor- between the estimate of one physician and that of
ders and predictors of human behavior. Much of the another regarding the advisability of tonsillectomy.
psychological research relevant to investment decision These and similar results prompted Bakwin to con-
making comes from the study of the judgmental pro- clude that, although the superstitions and magic rites
cesses of clinical psychologists, i.e.,  clinical judg- that prevailed in the 17th Century had largely been for-
ment. For a comprehensive review of research on clin- gotten, some theories and practices persisted in the
ical judgment see Goldberg [1968]. scientific era of medicine even though their falsity was
patent. Garland [1960] has provided similar and more
recent examples of unreliability and inaccuracy in
Accuracy and Reliability
medicine. For example, numerous studies of radiolo-
of Clinical Judgment
gists showed that they failed to recognize the presence
Over the past 20 years, numerous studies have tested of lung disease that was definitely visible on the X-ray
the accuracy of clinical judgments. The results, like film about 30 per cent of the time when reading X-rays.
those of investment judgments, have been quite discour- They also found that a radiologist changed his mind
aging. Goldberg references nine studies that not only about 20 per cent of the time when reading the same
demonstrated a marked lack of validity but also yielded film on two occasions.
the surprising finding that the judge s length of profes- This work in psychology and medicine, along with
sional training and his experience often showed little re- the previously described research in finance, implies
lationship to his accuracy. Equally disillusioning are the that we must never take for granted the reliability and ac-
15 experiments cited by Goldberg which showed that curacy of a judge, no matter how expert. Whenever pos-
162
PSYCHOLOGICAL STUDY OF HUMAN JUDGMENT
sible, empirical studies should be conducted to deter- Complex simulation. It is interesting that some
mine whether judgmental performance is satisfactory. of the most important analyses of complex judgment
processes were undertaken within the context of fi-
nancial decision making. Perhaps the outstanding ex-
ample is the work of Clarkson [1962] who undertook
Clinical vs. Statistical Prediction
to simulate the portfolio selection processes of a bank s
When one considers the typical findings of unre- trust investment officer. Clakson followed the officer
liability, lack of validity, and insensitivity to informa- around for several months and studied his verbalized
tion, it is not surprising to find clinical judgments in- reflections as he was asked to think aloud while re-
creasingly under attack by those who wish to substitute viewing past and present decisions. Using these verbal
statistical prediction systems for the human judge. Thus descriptions as a guide, the investment process was
psychology has its own version of the scientific vs. translated into a sequentially branching computer pro-
subjective controversy. This issue was popularized by gram. When the validity of the model was tested by
Meehl s classic book titled Clinical vs. Statistical Pre- comparing its selections with future portfolios selected
diction which was published in 1954 (Meehl, 1954). by the trust officer, the correspondence between actual
Goldberg [1968] summarized the vast amount of re- and simulated portfolios was fund to be remarkably
search stimulated by Meehl s book by pointing out that good. A similar research plan designed to simulate the
over a very diverse array of clinical tasks, some of which decision processes of bank officers when granting
were selected to show the clinician at his best and the business loans was outlined by Cohen, Gilmore, and
statistician at his worst, rather simple actuarial formulae Singer [1966].
typically performed at least as well as the clinical expert.
All this, of course, pertained to repetitive situations
Linear models. Clarkson s work shows that, giv-
where historical data existed for the statistician to use. It
en patient and intelligent effort, many of the expert s
would be interesting to pit clinical vs. statistical predic-
cognitions can be distilled into a form capable of being
tion methods in investment situations, although it is un-
simulated by a computer. However, there is still another
likely that the statistician s superiority would be any less
approach one that attempts to provide less of a se-
there than in the many studies in clinical psychology.
quential analysis and more of a quantified, descriptive
Recent studies by Sawyer [1966], Pankoff and Roberts
summary of the way that a decision maker weights and
[1968], and Einhorn [1972] indicate that a combination
combines information from diverse sources. This ap-
of clinical and statistical methods, with the clinician
proach aims to develop a mathematical model of the de-
gathering the data and the statistician processing it, may
cision maker and requires less time and effort on the part
be the optimal procedure to follow in many judgment
of investigator, subject, and computer. It forms a nice
situations.
compromise between Clarkson s complex, sequentially
branching model and the relatively naïve approaches of
the pre-computer era such as simply asking the deci-
Descriptive Analysis
sion maker how he makes his judgments. The rationale
of the Judgment Process
behind these mathematical models and techniques for
building them are reviewed by Slovic and Lichtenstein
Foresightful investment analysts have long recog-
[1971].
nized the need to understand more clearly the detailed
The basic approach requires the decision maker to
processes underlying investment decisions especial-
make quantitative evaluations of a fairly large number
ly decisions made by acknowledged experts. For ex-
of cases, each of which is defined by a number of quan-
ample, Bernhard [1959] observed that, if the metal
tified cue dimensions or characteristics. A financial an-
processes of consistently successful investors are intu-
alyst, for example, could be asked to predict the long-
itional, that intuitional reasoning must be made under-
term price appreciation for each of 50 securities, the se-
standable. In a similar vein, Bauman [1967] has argued
curities being defined in terms of cue factors such as
that by compelling the investment analyst to translate
their P/E ratios, corporate earnings growth trend, divi-
his vague attitudes, opinions, and reasons into explicit
dend yield, etc. Just as investigators interested in mod-
quantities, the analyst s thoughts are brought out into
eling the characteristics of the market have suggested
he open where they can be observed, evaluated, and
using multiple correlational procedures to capture the
tested.
way in which the market weights and responds to these
Researchers in the areas of economics, finance, and
factors, one could also fit a regression equation to the
psychology have recently taken up the challenge of
analyst s judgments to capture his personal weighting
simulating and describing the judgment process. At
policy. The resultant equation would be:
present, there are a number of new methods that should
be of interest to persons concerned with the dynamics
of investment decisions. Jpa = b1X1 + b2X2 + & bkXk (1)
163
SLOVIC
Where Jpa = predicted judgment of price apprecia- tween the volume of trading on a stock and its future
tions; X1, X2 & Xk are the quantitative values of the prospects:
defining cue factors (i.e., P/E ratios, earnings, etc.);
and b1, b2 & bk are the weights given to the various fac-  If you are driving a car you can get to your destina-
tion more quickly at 50 mph than at 10 mph. But you
tors in order to maximize the multiple correlation be-
may wreck the car at 100 mph. In a similar way, in-
tween the predicted judgments and the actual judg-
creasing volume on an advance up to a point is bullish
ments. These weights are assumed to reflect the rela-
and decreasing volume on a rally is bearish, but in
tive importance of the factors for the analyst. Equation
both cases only up to a point [Loeb, 1965; p. 287].
(1) is known as the linear model.
Psychologists have found linear models to be re-
Curvilinear functions such as this quote suggests can
markably successful in predicting judgments of such
be modeled by including exponential terms (i.e., x2i, x3i,
diverse phenomena as psychiatric diagnoses, malig-
etc.) as predictors in the judge s policy equation.
nancy of ulcers, job performance, and the riskiness and
When an analyst associates good investment deci-
attractiveness of gambles; and political scientists have
sions with complex and interrelated decision rules,
found linear models useful for describing judicial deci-
chances are that he envisages types of patterned or
sion processes in workman s compensation and civil
configural relationships rather than the linear combi-
liberties court cases (Slovic and Lichtenstein, 1971).
nation rule discussed above. Configurality means that
Researchers interested in simulating financial and
the analyst s interpretation of an item of information
managerial decisions have independently discovered
varies depending upon the nature of other available in-
the value of linear models. For example, Bowman
formation. This example of configural reasoning in-
[1963] and Kunreuther [1969] successfully fit linear
volving price changes, volume, and market cycle is
models to decisions concerned with production sched-
given by Loeb:
uling and Hester [1966] used regression analysis to
develop a  loan offer function representative of the
 Outstanding strength or weakness can have precisely
lending policy of a particular bank. Hester s function
opposite meanings at different times in the market cy-
makes explicit the weighting of such factors as the
cle. For example, consistent strength and volume in a
applicant s profits, his deposit balance, his current ratio
particular issue, occurring after a long general decline,
of assets to liabilities, etc. Such a function could be
will usually turn out to be an extremely bullish indica-
compared with the bank s formally-stated policy
tion. & On the other hand, after an extensive advance
guidelines. Functions of different loan officers could
which finally spreads to issues neglected all through the
also be compared.
bull market, belated individual strength and activity not
Large individual differences among weighting poli- only are likely to be shortlived but may actually suggest
theendofthegeneralrecovery& [Loeb,1965;p.65].
cies have been found in almost every study that reports
individual equations. A striking example of this in a
Since analysts believe that factors relevant to invest-
task demanding a high level of expertise comes from a
ment decisions should often be interpreted configur-
study of nine radiologists by Hoffman, Slovic, and
ally, it is important that techniques used to describe
Rorer [1968]. The stimuli were ulcers, described by the
judgment be sensitive to such processes. The linear
presence or absence of seven roentgenological signs.
model can be made sensitive to configural effects by
Each ulcer was rated according to its likelihood of
being malignant. There was considerable disagree- incorporating cross-product terms into the policy equa-
ment among the judgments of the radiologists, as indi- tion of the judge. Thus, if the meaning of factor X1 var-
cated by a median interjudge correlation, across stim- ies as a function of the level of factor X2, the term
uli, of only .38. Examination of each radiologist s lin- b12X1X2 can be added to the equation. A number of
studies have employed a statistical technique, analysis
ear equation clearly pinpointed the idiosyncratic
weightings of the various signs that lead to the ob- of variance, to identify configural processes in judg-
ment. For example, Slovic, Fleissner, and Bauman
served disagreements in diagnosis.
[1972] used this technique to isolate configural pro-
cesses used by stockbrokers when evaluating the at-
Nonlinear models. Although the linear model
tractiveness of common stocks. A number of interest-
does an impressive job of predicting judgments, when
one asks individuals how they are processing infor- ing instances of configural uses of information were
found in these studies and large differences in the pol-
mation their comments suggest that they use cues in a
variety of nonlinear ways. Researchers have attempt- icy equations for individual brokers were also evident.
ed to capture these nonlinear processes by means of
more complex equations. One type of nonlinearity
Subjective Weights and Self-Insight
occurs when an individual cue relates to the judg-
ments in a curvilinear manner. For example, this Thus far we have been discussing weighting poli-
quote from Loeb suggests a curvilinear relation be- cies that have been assessed by fitting an algebraic
164
PSYCHOLOGICAL STUDY OF HUMAN JUDGMENT
model to the judge s responses. We think of these as considerable empirical research. Other applications of
 computed or  objective policies. Judges in a num- bootstrapping have been described by Dawes [1971],
ber of studies were asked to estimate the relative Goldberg [1970], and Wiggins and Kohen [1971].
weights they had been using in the task. The corre-
spondence between these  subjective weights and
Studies of Probabilistic Inference
the computed weights indicated the judge s insight
into his own policy. One type of error in self-insight Conservatism. There is a rapidly developing
has emerged in all of these studies (Slovic and Lichten- school of thought called  Decision Theory which as-
stein, 1971). Judges strongly overestimate the impor- serts that we ought to cast our opinions about the
tance they place on minor cues (i.e., their subjective world in probabilistic terms [Edwards, Phillips, Hays
weights greatly exceed the computed weights for these and Goodman, 1968 and Raiffa, 1968]. For example,
cues) and they underestimate their reliance on a few rather than predicting that a stock will sell at a spe-
major variables. cific price six months from now, we should estimate a
In a recent study of 13 stockbrokers, Slovic, Fleis- probability distribution across a set of possible prices.
sner, and Bauman [1972] found an intriguing result These probabilities can then be used, in combination
that needs to be tested further. The longer a broker had with information about the payoffs associated with
been in the business, the less accurate was his insight various decisions and states of the world, to imple-
into his weighting policy. ment any of a number of decision rules, including the
maximization of expected value or expected utility.
When we translate our opinions into probabilities, a
Bootstrapping. Can a system be designed to aid
mathematical formula, Bayes theorem, dictates the
the decision maker that is based on his own judgments
optimal way that our estimates should change upon re-
of complex stimuli? One possibility is suggested by the
ceipt of new information. Led by the efforts of Ed-
finding that algebraic models, such as the linear model,
wards [1965, 1968], many psychologists have com-
can do a remarkably good job of simulating such judg-
pared man s subjective probability revisions with those
ments. An important hypothesis about cooperative in-
of Bayes theorem in a variety of experimental and
teraction between man and machine is that the model
real-life situations. This research shows that men are
of the man may be able to make better predictions than
conservative processors of fallible information. Upon
the man himself. Dawes [1971] has termed this phe-
receipt of new data, subjects revise their probability es-
nomenon  bootstrapping.
timates in the direction prescribed by Bayes theorem,
The rationale behind the bootstrapping hypothesis
but the revision is typically too small; subjects respond
is quite simple. Although the human judge possesses
as though the data are less diagnostic than they truly
his full share of human learning and hypothesis-gener-
are. In some studies subjects have required from two to
ating skills, he lacks the reliability of a machine. As
nine data observations to revise their estimates as much
Goldberg [1970] has noted:
as Bayes theorem would prescribe after just one obser-
vation. A number of experiments have attempted to ex-
 He is subject to all these human frailties which lower
plain this finding. The results are controversial but in
the reliability of his judgments below unity. And, if the
Edwards view [1968] the major cause of conservatism
judge s reliability is less than unity, there must be error
is human misaggregation of the data. That is, men per-
in his judgments error which can serve no other pur-
ceive a datum accurately and are well aware of its indi-
pose than to attenuate his accuracy. If we could &
[eliminate] the random error in his judgments, we vidual diagnostic meaning, but are unable to combine
should thereby increase the validity of the resulting
its meaning properly with their prior opinions and with
predictions (Goldberg, 1970; p. 423).
the diagnostic meaning of other data when revising
their estimates.
The algebraic model captures the judge s weighting
policy and applies it consistently. If there is some va- Intuitions about sampling variability. There is
lidity to this policy to begin with, filtering out the error a different type of probabilistic inference in which de-
via the model should increase accuracy. Of course, cision makers turn out to be very nonconservative. This
bootstrapping preserves and reinforces any miscon- work is described by Tversky and Kahneman [1971]
ceptions or biases that the judge may have. Implicit in who analyzed the decisions psychologists make when
the use of bootstrapping is the assumption that these bi- planning their scientific experiments. Despite formal
ases will be less detrimental to performance than the training in statistics, psychologists usually rely upon
inconsistencies of unaided human judgment. their educated intuitions when they decide how large a
Bootstrapping has been explored independently by sample of data to collect or whether they should repeat
a number of different investigators. Bowman [1963] an experiment to make sure their results are reliable.
outlined a bootstrapping approach within the context Tversky and Kahneman distributed a questionnaire to
of managerial decision making that has stimulated psychologists in the audience at the meetings of several
165
SLOVIC
professional societies. Typical of the questions was the other answers involve some version of the decomposi-
following: tion principle:
 Suppose you have run an experiment on 20 subjects,  The spirit of decision analysis is divide and conquer:
and have obtained a significant result which confirms Decompose a complex problem into simpler prob-
your theory (z = 2.23, p < .05, two-tailed). You now lems, get your thinking straight in these simpler prob-
have cause to run an additional group of 10 subjects. lems, paste these analyses together with a logical glue,
What do you think the probability is that the results and come out with a program for action for the com-
will be significant, by a one-tailed test, separately for plex problem. Experts are not asked complicated, fuz-
this group [p. 105]? zy equations, but crystal clear, unambiguous, elemen-
tal, hypothetical questions [Raiffa, 1968; p. 271].
From the answers to this and a variety of other ques-
tions, Tversky and Kahneman concluded that people There seems to be general agreement that we cannot
have strong intuitions about random sampling; that do away with the human element in judgment, so the
these intuitions are wrong in fundamental ways; that decomposition approach attempts to obtain relatively
they are shared by naïve persons and sophisticated sci- simple judgments that can be integrated by some opti-
entists alike; and that they are applied with unfortunate mal combination model. In this way, it relieves the
consequences in the course of scientific inquiry. They judge from having to integrate his basic opinions and
found that the typical scientist gambles his research expectations. For example, certain information-pro-
hypotheses on small samples without realizing that the cessing systems require men to estimate the prob-
odds against his obtaining accurate results are unrea- ability that each of various items of data would be
sonably high; has undue confidence in early trends observed, given a certain state of the world. These es-
from the first few data points and in the stability of ob- timates are then processed mechanically via Bayes
served patterns; has unreasonably high expectations theorem to produce an estimate of the probability of
about the replicability of significant results; and rarely that state, given that this data was observed (Edwards,
attributes a deviation of results from expectations to Lindman and Phillips, 1968; Edwards, Phillips, Hays
sampling variability because he finds a causal explana- and Goodman, 1968). In the realm of finance, portfolio
tion for any discrepancy. selection models require that analysts estimate expect-
Tversky and Kahneman summarized these results ed returns, variances, covariances, etc., which are then
by asserting that people s intuitions seemed to satisfy a combined via an optimal model (Ahlers, 1966 and
 law of small numbers which means that the  law of Markowitz, 1959). Similarly, models of common stock
large numbers applies to small samples as well as to valuation require an analyst to make estimates of future
large ones. The  law of large numbers says that very balance-sheet and income data, which can be com-
large samples will be highly representative of the pop- bined by an empirically-derived or theoretically-based
ulation from which they are drawn. Thus, small sam- model (Ahlers, 1966, Wendt, 1965). The use of deci-
ples were also expected to be highly representative of sion trees to analyze complex investment problems is
the population. Since acquaintance with logic or prob- another example of decomposition (Magee, 1964 and
ability theory did not make the scientist any less sus- Raiffa, 1968).
ceptible to these cognitive biases, Tversky and Kah- Decomposition is certainly a reasonable approach,
neman concluded that the only effective precaution is although it is still too early to know how successful it
the use of formal statistical procedures, rather than in- will be. Critics claim that the decision maker may be
tuition, to design experiments and evaluate data. able to make good judgments and choices without
In a related study of college undergraduates, Kah- being able to introspect accurately about the values
neman and Tversky [1972] found that many of these and expectations that underlie his actions. A decision
individuals did not understand that fundamental princi- maker who has developed an expertise in a particular
ple of sampling, i.e., the notion that the error in a sam- area may find it extremely difficult and unnatural to re-
ple becomes smaller as the sample size gets larger. spond to elemental questions about which he has never
Kahneman and Tversky concluded:  For anyone who thought and with which he has had no direct experi-
would wish to view man as a reasonable intuitive stat- ence. In addition, there are a number of biases that dis-
istician, such results are discouraging. tort even the simplest kinds of judgments of probabil-
ity, variance, or correlation, as the remainder of this
section will illustrate.
Biases in Judgments of Probability,
Variability, and Covariation
Biased judgments of probability. Some of the
What can be done to help the decision maker inter- inadequacies of probabilistic judgment have already
pret and combine information appropriately? Boot- been discussed. In addition to conservatism and the
strapping is one answer to this question. Most of the belief in the law of small numbers, there is yet an-
166
PSYCHOLOGICAL STUDY OF HUMAN JUDGMENT
other source of distortion,  availability bias, that person to person. Some people are overly optimistic,
affects simple probability estimates. According to tending to attribute greater probability to highly-de-
Tversky and Kahneman [1971], the essence of this sired events than to undesired events, other factors be-
bias is that judgments of an event s probability are ing equal. Other persons consistently overestimate the
determined by the number of instances of that event likelihood of unpleasant events.
that are remembered and the ease with which they
come to mind. The availability of instances is af- Biased judgments of variance. Several factors
fected by such factors as recency, salience, and imag- seem to influence a person s judgment of the variance
inability, all of which may or may not be related to of a sequence of values about the mean of that se-
the correct probability. For example, the letter k is quence (Beach and Scopp, 1968 and Lathrop, 1967).
three times as likely to appear as the third letter of an The first of these factors is the mean itself. Perceived
English word as the first letter, yet most persons variance increases as the mean decreases. A standard
judge it as more likely to be a first letter. Tversky and deviation of two feet for a group of saplings would be
Kahneman hypothesize that, when subjects make this perceived as larger than the same standard deviation
judgment, they try to think of words either beginning for a group of fully-grown trees. Greater irregularity in
with k or having k as the third letter. It is easier to a sequence also leads to an illusion of greater variance.
think of words that begin with k, and if we use that Sequences in which the values progress in an orderly
fact as a cue on which to base our intuitive probabil- fashion (e.g., ascending or descending or ascending up
ity estimates, these words will be perceived as more to a point, then descending, etc.), with little difference
probable than words with k in the third position. In between successive values, are perceived to have less
general, the harder it is to recall or imagine instances variance than sequences whose adjacent values are less
of an event, the lower the judged probability of that regular (Lathrop, 1967). The stimuli in experiments on
event. perceived variance have been sets of line lengths and
The effects of availability bias are not likely to be numbers. It would be interesting to determine whether
limited to the psychological laboratory. An analyst these same sorts of biases would occur when the vari-
who attempts to evaluate the likelihood of a recession ance of a sequence of stock prices or earnings reports
may do so by recalling economic conditions similar to was being judged.
those of the present or by recalling recessions. The lat-
ter are easier to retrieve because they are more sharply Biased judgments of correlation and causal-
defined, whereas states of the economy are more dif- ity. There have been a number of studies relating to
ficult to characterize and, therefore, harder to remem- judgments of correlation and causality. The results of
ber. The resulting probability estimate is likely to be these studiessuggest thatevenif therandomwalktheory
greatly dependent upon which of these two mental sets of security price changes were absolutely true, we prob-
the analyst adopts. Even the form of the question may ably would not believe it and would find, upon ob-
be important. Consider the following questions: serving random price changes, what appear to be mean-
ingful patterns upon which to base our forecasts.
a)  How likely is it that there will be a recession Several lines of psychological research appear rele-
soon? vant here. The first stems from a classic experiment by
b)  How likely is it that, with the present tighten- Skinner [1948]. Skinner found that hungry birds, given
ing of credit, there will be a recession soon? food at brief random intervals, developed very idiosyn-
cratic, repetitive actions. The precise form of this be-
The first question may focus attention on past in- havior varied from bird to bird, and Skinner referred to
stances of recession, whereas the latter may cause the these actions as superstitions. What happened to these
analyst to think about previous credit conditions. birds can be described in terms of the concept of posi-
There are numerous other instances of systematic tive reinforcement. The delivery of food increased the
biases in our judgments of probabilities. Cohen and likelihood of whatever form of behavior happened to
Chesnick [1970] and Slovic [1969] found that subjects precede it. Food was then presented again. Because the
systematically misperceived the probabilities of com- reinforced behavior was occurring at an increased rate,
pound events. For example, in the study by Cohen and it was more likely to be reinforced again. The second
Chesnick, some people preferred the opportunity of reinforcement caused a further increase in the rate of
drawing a winning lottery ticket out of a population of this particular behavior which improved its chances of
10 tickets (with one attempt) to the chance to draw the being reinforced again, and so on. After a short while
winner out of 100 tickets, even when they had up to the birds were found to be turning rapidly counter
twenty draws of the latter kind (with replacement after clockwise about the cage, hopping from side to side,
each draw). Other studies have found that the desirabil- making odd head movements, etc. Because such be-
ity of an event biases its subjective probability (Slovic, haviors are reinforced less than 100 per cent of the time
1966) although the effects are complex and differ from during learning, they persist even when reinforcement
167
SLOVIC
stops altogether. Animals trained in this way have been is that, if we provide a stream of random price
known to make as many as 10,000 attempts to obtain a changes to intelligent but naïve subjects, say under-
reward that was no longer forthcoming. graduate students in a finance course, they might dis-
The environment of the stock market seems to pro- cover in these random sequences some of the same
vide exactly the right conditions for the development rules that we see accepted by chartists or other ana-
and maintenance of superstitious behavior. That is, lysts. Although the influence of illusory correlation in
there has been a favorable expected return and thus a financial analysis remains to be demonstrated, there
predominance of positive reinforcement (at least in the is no reason to believe that it will be less than in clini-
past) which is administered intermittently. And there is cal psychology.
always the hope that if enough people harbor the same Finally, a number of studies have investigated sub-
superstitions, and the game is to anticipate the actions jects perceptions of correlation and causality in simple
of the crowd, then knowing the superstitions and acting situations involving just two binary variables. Consider
on them may be quite rewarding. At any rate, one a 2 × 2 table in which variable A is the antecedent or in-
chartist may have been correct when he said,  If I put variable and B is the consequent or output variable;
hadn t made money some of the time, I would have ac- and the small letters are the joint frequencies.
quired market wisdom quicker (Lefevre, 1968; p. 30).
The superstitions developed in Skinner s pigeons B1 B2
were highly individualistic. Yet the behavior and the
A1 a b
lore of Wall Street is often commonly agreed upon.
How can this consensus be reconciled with the notion
A2 c d
of stock-price changes as a random walk? Several re-
cent experiments by Chapman and Chapman [1967]
A correlation or contingency exists between A and
may provide a possible answer to this question along
B to the extent that the probability of B1 given A1 dif-
with further insight into the pitfalls awaiting human
fers from the probability of B1 given A2: that is, to the
intuition.
extent that a/(a + b) differs from c/(c+ d).
The Chapmans, studying a phenomenon they have
Research indicates that subjects judgments of con-
labeled illusory correlation, have shown how our prior
tingency are not based on a comparison of a/(a + b) ver-
expectations of relationships can lead to faulty obser-
sus c/(c + d). For example, Smedslund [1963] had stu-
vation and inference, even under seemingly excellent
dents of nursing judge the relation between a symptom
conditions for learning. They presented naïve subjects
and the diagnosis of a disease. He found that the judg-
with human figure drawings, each of which was paired
ments were based mainly on the frequency of joint oc-
with a statement about the personality of the patients
currence of symptom and disease (cell a in the table),
who allegedly drew the figures. These statements were
without taking the other three event combinations into
randomly paired with the figure drawings so that the
account. As a result, the judgments were unrelated to
figure cues were unrelated to the personality of the
actual contingency. Similar results were obtained by
drawer. They found that most subjects learned to see
Ward and Jenkins [1965] who concluded:
what they expected to see. In fact, naïve subjects dis-
covered the same relationships between drawings and
 In general & statistically naïve subjects lack an ab-
personality that expert psychologists report observing
stract concept of contingency that is isomorphic with
in their clinical practice, although these relationships
the statistical concept. Those who receive information
were absent in the experimental materials. The illusory
on a trial basis, as it usually occurs in the real world,
correlates corresponded to commonly-held expecta-
generally fail to assess adequately the degree of rela-
tions, such as figures with big eyes being drawn by sus- tionship present [Ward and Jenkins, 1965; p. 240].
picious people, muscular figures being drawn by indi-
viduals who worried about their manliness, etc.
The Chapmans noted that in clinical practice the Experimental Study
observer is reinforced in his observation of illusory cor- of Risk-Taking Behavior
relates by the reports of his fellow clinicians, who them-
selves are subject to the same illusions. Such agreement There is a great deal of experimental research on
among experts is, unfortunately, often mistaken as evi- risk-taking behavior that may have implications for in-
dence for the truth of the observation. They concluded vestment decision making. In this research, subjects
that the clinician s cognitive task may exceed the capac- are asked to indicate their preferences and opinions
ityof the human intellect andthey suggestedthat subjec- among various gambles. Gambles are studied because
tive intuition may need to be replaced, at least partially, they represent, in abstract form, important aspects of
by statistical methods of prediction. real-life decisions namely, probabilities, incentives,
The research on illusory correlation suggests par- and risks. By using gambles, the basic dimensions of
allel experiments using stock prices. One hypothesis risk-taking situations can be manipulated and hypothe-
168
PSYCHOLOGICAL STUDY OF HUMAN JUDGMENT
ses can be tested in a rigorous way. Whether the results they would like to play each bet in a pair. First they
generalize to real-life gambles must, of course, be made a simple choice, A or B. Later they were asked to
checked by further research. assume they owned a ticket to play each bet, and they
were to state the lowest price for which they would sell
this ticket.
The Influence of Variance
Presumably these selling prices and choices are
on Risk Taking
both governed by the same underlying quality, the
Theorists such as Allais [1953], Fisher [Fisher], and subjective attractiveness of each gamble. Therefore,
Markowitz [1959] have argued that the variance of re- the subject should state a higher selling price for the
turns on an investment should be considered as an in- gamble that he prefers in the choice situation. How-
vestment criterion in addition to the mean or expected ever, Lichtenstein and Slovic found that subjects often
return. High variance is typically equated with high chose Bet A, yet stated a higher selling price for Bet B.
risk. Why should this happen? Lichtenstein and Slovic have
Does variance influence the perceived attractive- traced it to the fact that subjects used different cog-
ness of a gamble? Subjects in several psychological ex- nitive strategies for setting prices than for making
periments have exhibited what seemed to be strong choices. Subjects choose Bet A because of its good
preferences for playing high or low variance gambles odds, but they set a higher price for B because of its
(for example, see Coombs and Pruitt, 1960). However, large winning payoff.
recent evidence suggests that the subjects in these ex- A  compatibility effect seemed to be operating
periments were choosing according to decision rules here. Since a selling price is expressed in terms of mon-
such as  minimize possible loss or  maximize possi- etary units, subjects apparently found it easier to use
ble gain, rather than basing their preferences on vari- the monetary aspects of the gamble to produce this
ance per se. Variance appears to have correlated with type of response. Such a bias did not exist with the
the preferences only because it also correlated with choices. When faced with their inconsistent decisions,
these other strategies (Slovic and Lichtenstein, 1968). many subjects had a very hard time changing either of
Another study has found that perceived risk was not their conflicting responses. They felt that the different
a function of the variance of a gamble (Slovic, 1967). strategies they used for each decision were appropriate.
Instead, riskiness was more likely to be determined by However, strict adherence to an inconsistent pattern of
the probability of loss and the amount of loss. This re- prices and choices can be termed irrational, since the
sult is in accord with comments made by Lorie [1966] inconsistent subject can be led into purchasing and
who complained that it was absurd to call a stock risky trading gambles in such a way that he continually loses
because it went up much faster than the market in some money.
years and only as fast in other years, while a security The message in this research is that integrating in-
that never varies in price is not risky at all, if variance is formation is quite a difficult cognitive task, and there
used to define risk. The importance of understanding may often be a very subtle interaction between the
how risk is perceived is stressed by Lepper [1967], who form of the information we have to use and the form of
pointed out the crucial role of investors perceptions of the judgmental response we have to make. This may
risk in determining the impact of various taxes. Taxes, well generalize beyond experimental gambling situa-
of course, can alter markedly the variance of the poten- tions. For example, a financial analyst who is forecast-
tial returns for an investment. ing a stock s market price six months hence might be
led to overweight previous price information, simply
because of the compatibility factor. And if he were
Response Mode and Information Use
asked to forecast percentage price increase rather than
A study by Lichtenstein and Slovic [1971] found price itself, he might then give more weight to other
that subtle changes in the manner in which the decision variables in the company report that were expressed in
maker reported his evaluation of a gamble had a strong terms of percentages. Experiments testing this hypoth-
influence on the way that he processed information esis would seem to be worth conducting, so that steps
about probabilities and payoffs. For example, consider could be taken to minimize compatibility biases if they
the following pair of bets: are found.
Bet A: .90 to win $4 and .10 to lose $2
Is Willingness to Take Risks a Stable
Bet B: .30 to win $16 and .70 to lose $2.
Personality Trait?
Bet A has a much better probability of winning but An understanding of risk-taking propensity as a per-
Bet B offers a higher winning payoff. Lichtenstein and sonality characteristic could prove valuable in the se-
Slovic s subjects were shown many such pairs of bets. lection and training of portfolio managers, investment
They were asked to indicate, in two ways, how much counselors, or brokers. It would also help these indi-
169
SLOVIC
viduals to understand and service their clients better. themselves as being at least as willing as their peers to
Although knowledge of the dynamics of risk taking is take risks. In this regard, the group discussion provides
still limited, there is one important aspect that has been information that allows group members to compare
fairly well researched that dealing with the stability their own positions with those of their peers. Members
of a person s characteristic risk-taking preferences as whose initial positions were less risky than those of the
he moves from situation to situation. Typically, a sub- group average come to learn that they are not as risky
ject is tested in a variety of risk-taking tasks involving as they thought and as they want to be. To remedy this,
problem solving, athletic, social, vocational, and pure they increase their level of risk taking.
gambling situations. The results of close to a dozen Both of these explanations, and others as well, have
such studies indicate little correlation, from one setting received experimental verification. For more detailed
to another, in a person s preferred level of risk taking discussion of group influence on risk taking see re-
(Slovic, 1972). Only those tasks highly similar in struc- views by Clark [1971] and Kogan and Wallach [1967].
ture and involving the same sorts of payoffs (e.g., all fi-
nancial, all social, etc.) have shown any generality and,
as similarity decreases, these cross-task consistencies
Concluding Remarks
rapidly decline. Thus an individual who takes risks by
guessing often on a mathematics exam (when guessing
Several facts are important about the research de-
is penalized) is likely to be a high risk taker in other
scribed in this paper. First, most of the work is of very
exams as well, but that does not imply that he would
recent origin. Second, with only a few exceptions, this
prefer a high-risk occupation. In sum, the majority of
research has been done without explicit consideration
evidence argues against the existence of risk-taking
of problems in business and finance. As a result, there
propensity as a generalized characteristic of individu-
is a great need to replicate the various types of studies
als. A person s previous learning experiences in spe-
in specific financial settings. Studies of high-level de-
cific risk-taking settings seem much more important
cision makers and analysts, in their natural working
than his general personality characteristics.
environment, are particularly needed. Besides contrib-
As an example of one implication of this work, con-
uting to the understanding of financial decisions, such
sider the problem of selecting a portfolio manager.
research would also benefit psychology, much as Clark-
Suppose that one desires a manager who has the pro-
son s simulation of the trust investment officer pro-
pensity to invest at high levels or risk. The best predic-
vided important insight into the nature of complex
tor of this characteristic would be the individual s dem-
thought processes. Obviously, this kind of research
onstrated performance in a position highly similar to
would benefit greatly from interdisciplinary collabora-
the one under consideration. Evidence of his risk-tak-
tion among psychologists, economists, financial ana-
ing propensity gleaned from other forms of behaviors
lysts, computer scientists, and others.
is unlikely to predict how he would behave in an invest-
If research in financial settings verifies the early in-
ment situation.
dications of man s information-processing limitations,
the next phase of research must certainly emphasize
the development of techniques to help decision makers
Comparison of Group
overcome their cognitive biases. Will informing an an-
and Individual Risk Taking
alyst about his biases make him less susceptible to
Many decisions are made not by individuals, but by
them or will it lead him to overcompensate, perhaps
groups. Over the past decade, comparison of group
with even greater error? Would computer simulation
versus individual risk-taking tendencies has been the
be effective in conveying an appreciation of sampling
subject of an extensive body of research. The typical
variability and probabilities? The past decade of re-
finding is that decisions made by groups are riskier
search has uncovered some fascinating questions. The
than that average of the individual members decisions
next decade should provide some extremely interesting
prior to group discussion. Individual risk-taking levels
and important answers.
also increase following group discussion. This phe-
nomenon has been labeled the  risky shift.
One of the leading explanations of the risky shift is
References
the  diffusion of responsibility hypothesis. It asserts
that each group member feels less personal blame if his
Ahlers, D. M.  SEM: A Security Evaluation Model, in K. J. Cohen
choice fails, thus he is not afraid to recommend or ac- and F. S. Hammer (Eds.), Analytical Methods in Banking.
Homewood, Ill.: R. D. Irwin, 1966, pp. 305 336.
cept riskier courses of action.
Allais, M.  Le Comportement de l homme Rationnel Devant le Ris-
Another explanation of the risky shift is the  cul-
que. Critique des Postulate et Axioms de l école Americanine,
tural value hypothesis which assumes that moderate
Econometrica, 21, (1953), pp. 503 546.
riskiness is stronger, more widely held cultural value
Bakwin, H.  Pseudodoxia Pediatrica, New England Journal of
than caution. This value leads individuals to perceive Medicine, 232, (1945), pp. 691 697.
170
PSYCHOLOGICAL STUDY OF HUMAN JUDGMENT
Bauman, W. S.  The Less Popular Stocks Versus the Most Popular Graham, B., D. L. Dodd, S. Cottle and C. Tatham. Security Analy-
Stocks, Financial Analysts Journal, 21, (1965), pp. 61 69. sis: Principles and Technique. New York: McGraw-Hill,
Bauman, W. S.  Scientific Investment Analysis, Financial Analysts 1962.
Journal, 23, (1967), pp. 93 97. Gray, W. S.  Measuring the Analyst s Performance, Financial Ana-
Beach, L. R. and T. S. Scopp.  Intuitive Statistical Inferences About lysts Journal, 22, (1966), pp. 56 60.
Variances, Organizational Behavior and Human Performance, Hausman, W. H.  A Note on  The Value Line Contest: A Test of the
3, (1968), pp. 109 123. Predictability of Stock-Price Changes,  Journal of Business,
Bernhard, A. The Evaluation of Common Stocks. New York: Simon 42, (1969), pp. 317 320.
and Schuster, 1959. Hester, D. D.  An Empirical Examination of a Commercial Bank
Bernhard, A.  The Results of the Second (1967) Value Line Contest Loan Offer Function, in K. J. Cohen & F. S. Hammer (Eds.),
in Stock Market Judgment, The Value Line Investments Survey, Analytical Methods in Banking. Homewood, Ill.: R. D. Irwin,
23, (1967), pp. 236 242. 1966, pp. 178 217.
Bowman, E. H.  Consistency and Optimality in Managerial Deci- Hoffman, P. J., P. Slovic and L. G. Rorer.  An Analysis-of-Variance
sion-Making, Management Science, 9, (1963), pp. 310 321. Model for the Assessment of Configural Cue Utilization in
Chapman, L. J. and J. P. Chapman.  Genesis of Popular But Errone- Clinical Judgment, Psychological Bulletin, 69, (1968), pp.
ous Psychodiagnostic Observations, Journal of Abnormal Psy- 338 349.
chology, 72, (1967), pp. 193 204. Jensen, M. C.  The Performance of Mutual Funds in the Period
Clark, III, R. D.  Group-Induced Shift Toward Risk: A Critical Ap- 1945 1964, Journal of Finance, 23, (1968), pp. 389 416.
praisal, Psychological Bulletin, 76, (1971), pp. 251 270. Kahneman, D. and A. Tversky.  Subjective Probability: A Judgment
Clarkson, G. P. E. Portfolio Selection: A Simulation of Trust Invest- of Representativeness, Cognitive Psychology, 3, (1972), pp.
ment. Englewood Cliffs, NJ: Prentice-Hall, 1962. 430 454.
Cohen, J. and E. I. Chesnick.  The Doctrine of Psychological Kogan, N. and M. A. Wallach.  Risk Taking as a Function of the Sit-
Chances, British Journal of Psychology, 61, (1970), pp. uation, the Person, and the Group, New Directions in Psychol-
323 334. ogy: III. New York: Holt, Rinehart, & Winston, 1967, pp.
Cohen, K. J., T. C. Gilmore and F. A. Singer.  Bank Procedures for 111 278.
Analyzing Business Loan Applications, in K. J. Cohen & F. S. Kunreuther, H.  Extensions of Bowman s Theory on Managerial De-
Hammer (Eds.), Analytical Methods in Banking. Homewood, cision Making. Management Science, 15, (1969), pp. 415
Ill.: R. D. Irwin, 1966, pp. 218 251. 439.
Coombs, C. H. and D. G. Pruitt.  Components of Risk in Decision Lathrop, R. G.  Perceived Variability, Journal of Experimental Psy-
Making: Probability and Variance Preferences, Journal of Ex- chology, 73, (1967), pp. 498 502.
perimental Psychology, 60, (1960), pp. 265 277. Lefevre, E. Reminiscences of a Stock Operator. New York: Pocket
Cowles, A.  Can Stock Market Forecasters Forecast? Economet- Books, 1968.
rica, 1, (1933), pp. 309 324. Lepper, S. J.  Effects of Alternative Tax Structures on Individuals
Cowles, A.  Stock Market Forecasting, Econometrica, 12, (1944), Holdings of Financial Assets, in D. D. Hester & J. Tobin
pp. 206 214. (Eds.), Risk Aversion and Portfolio Choice. New York: Wiley,
Cragg, J. G. and B. G. Malkiel.  The Consensus and Accuracy of 1967, pp. 51 109.
Some Predictions of the Growth of Corporate Earnings, Jour- Lichtenstein, S. and P. Slovic.  Reversals of Preference Between
nal of Finance, 23, (1968), pp. 67 84. Bids and Choices in Gambling Decisions, Journal of Experi-
Dawes, R. M.  A Case Study of Graduate Admissions: Application mental Psychology, 89, (1971), pp. 46 55.
of Three Principles of Human Decision Making, American Loeb, G. The Battle for Investment Survival. New York: Simon and
Psychologist, 26, (1971), pp. 180 188. Schuster, 1965.
Drew, G. A. New Methods for Profit in the Stock Market. Boston: Lorie, J. H.  Some Comments on Recent Quantitative and Formal
Metcalf Press, 1941. Research on the Stock Market, Journal of Business, 39, Part II,
Edwards, W.  Conservatism in Human Information Processing, in (1966), pp. 107 110.
B. Kleinmuntz (Ed.), Formal Representation of Human Judg- Magee, J. F.  Decision Trees for Decision Making, Harvard Busi-
ment. New York: Wiley, 1968, pp. 17 52. ness Review, 42, (1964), pp. 126 138.
Edwards, W., H. Lindman and L. D. Phillips.  Emerging Technol- Markowitz, H. M. Portfolio Selection: Efficient Diversification of In-
ogies for Making Decisions, New Directions in Psychology: II. vestments. New York: Wiley, 1959.
New York: Holt, Rinehart, & Winston, 1965, pp. 261 325. Meehl, P. E. Clinical Versus Statistical Prediction: A theoretical
Edwards, W., L. D. Phillips, W. L. Hays and B. C. Goodman.  Proba- Analysis and a Review of the Evidence. Minneapolis: Univer-
bilistic Information Processing Systems: Design and Evalua- sity of Minnesota Press, 1954.
tion, IEEE Transactions on Systems Science and Cybernetics, Murphy, J. M.  The Value Line Contest: 1969, Financial Analysts
SSC-4, (1968), pp. 248 265. Journal, 26, (1970), pp. 94 100.
Einhorn, H. J.,  Expert Measurement and Mechanical Combina- O Brien, J. W.  How Market Theory Can Help Investors Set Goals,
tion, Organizational Behavior and Human Performance, 7, Select Investment Managers and Appraise Investment Perfor-
(1972), pp. 86 106. mance, Financial Analysts Journal, 26, (July Aug., 1970), pp.
Fama, E. F.  The Behavior of Stock Market Prices, Journal of Busi- 91 103.
ness, 38, (1965), pp. 34 105. Oskamp, S.  Overconfidence in Case-Study Judgments, Journal of
Fisher, I. The Nature of Capital and Income. New York: Macmillan, Consulting Psychology, 29, (1965), pp. 261 265.
1906. Pankoff, L. D. and H. V. Roberts.  Bayesian Synthesis of Clinical
Garland, L. H.  The Problem of Observer Error. Bulletin of the New and Statistical Prediction, Psychological Bulletin, 70, (1968),
York Academy of Medicine, 36, (1960), pp. 569 584. pp. 762 773.
Goldberg, L. R.  Simple Models or Simple Processes? Some Re- Raiffa, H. Decision Analysis. Reading, Mass.: Addison-Wesley,
search on Clinical Judgments, American Psychologist, 23, 1968.
(1968), pp. 483 496. Sawyer, J.  Measurement and Prediction, Clinical and Statistical,
Goldberg, L. R.  Man Versus Model of Man: A Rationale, Plus Psychological Bulletin, 66, (1966), pp. 178 200.
Some Evidence, for a Method of Improving on Clinical Infer- Skinner, B. F.  Superstition in the Pigeon, Journal of Experimental
ences, Psychological Bulletin, 73, (1970), pp. 422 432. Psychology, 38, (1948), pp. 168 172.
171
SLOVIC
Slovic, P.  Value as a Determiner of Subjective Probability, IEEE Smedslund, J.  The Concept of Correlation in Adults, Scandinavian
Transactions on Human Factors in Electronics, HFE-7, (1966), Journal of Psychology, 4, (1963), pp. 165 173.
pp. 22 28. Smith, A. The Money Game. New York: Random House, 1968.
Slovic, P.  The Relative Influence of Probabilities and Payoffs Upon Treynor, J. L. and K. K. Mazuy.  Can Mutual Funds Outguess the
Perceived Risk of A Gamble, Psychonomic Science, 9, (1967), Market? Harvard Business Review, 44, (1966), pp. 131
pp. 223 224. 136.
Slovic, P.  Manipulating the Attractiveness of a Gamble Without Tversky, A. and D. Kahneman.  Belief in the Law of Small Num-
Changing Its Expected Value, Journal of Experimental Psy- bers, Psychological Bulletin, 76, (1971), pp. 105 110.
chology, 79, (1969), pp. 139 145. Tversky, A. and D. Kahneman.  The Judgment of Frequency and
Slovic, P.  Information Processing, Situation Specificity, and the Probability by Availability of Instances, Oregon Research In-
Generality of Risk-Taking Behavior, Journal of Personality stitute Research Bulletin, Vol. 11, No. 6, (1971).
and Social Psychology, 22, (1972), pp. 128 134. Ward, W. C. and H. Mm Jenkins.  The Display of Information and
Slovic, P., D. Fleissner and W. S. Bauman.  Analyzing the Use of In- the Judgment of Contingency, Canadian Journal of Psychol-
formation in Investment Decision Making: A Methodological ogy, 19, (1965), pp. 231 241.
Proposal, Journal of Business, 45, (1972), pp. 283 301. Wendt, P. F.  Current Growth Stock Valuation Methods, Financial
Slovic, P. and S. Lichtenstein.  The Importance of Variance Prefer- Analysts Journal, 21, (1965), pp. 91 103.
ences in Gambling Decisions, Journal of Experimental Psy- Wharton School of Finance and Commerce.  A Study of Mutual
chology, 78, (1968), pp. 646 654. Funds, Report of the Committee on Interstate and Foreign
Slovic, P. and S. Lichtenstein.  Comparison of Bayesian and Regres- Commerce, 87th Congress, House Report 2274, 1962.
sion Approaches to the Study of Information Processing in Wiggins, N. and E. Kohen.  Man vs. Model of Man Revisited: The
Judgment, Organizational Behavior and Human Performance, Forecasting of Graduate School Success, Journal of Personal-
6, (1971), pp. 649 744. ity and Social Psychology, 19, (1971), pp. 100 106.
172