Journal ol Experimental Psychology:

Human Perception and Performance

1975, Vol. 1, No. 3, 288-299

Hindsight ^Foresight: The Effect of Outcome Knowledge

on Judgment Under Uncertainty
Baruch Fischhoff
Hebrew University of Jerusalem, Israel
One major difference between historical and nonhistorical judgment is that
the historical judge typically knows how things turned out. In Experiment
1, receipt of such outcome knowledge was found to increase the postdicted
likelihood of reported events and change the perceived relevance of event-
descriptive data, regardless of the likelihood of the outcome and the truth
of the report. Judges were, however, largely unaware of the effect that
outcome knowledge had on their perceptions. As a result, they overesti-
mated what they would have known without outcome knowledge (Experi-
ment 2), as well as what others (Experiment 3) actually did know without
outcome knowledge. It is argued that this lack of awareness can seriously
restrict one's ability to judge or learn from the past.

Hindsight and foresight differ formally in Reporting an outcome's occurrence increases

the information available to the observer. its perceived probability of occurrence; and
The hindsightful judge possesses outcome (b) people who have received outcome
knowledge, that is, he knows how things knowledge are largely unaware of its having
turned out. The foresightful judge does changed their perceptions in the manner de-
not. Although foresight usually implies scribed in the first hypothesis. In combina-
looking at the future, in the absence of out- tion, these two hypotheses indicate that re-
come knowledge, past and future events can porting an outcome produces an unjustified
be equally inscrutable. increase in its perceived predictability, for it
The studies presented here ask two ques- seems to have appeared more likely than it
tions about the judgmental differences be- actually was.
tween hindsight and foresight: (a) How Indirect support for the first hypothesis
does receipt of outcome knowledge affect may be found in a variety of sources. For
judgment? (b) How aware are people of the example, the historian Georges Florovsky
effects that outcome knowledge has on their (1969) notes:
perceptions? Answers to these questions The tendency toward determinism is somehow im-
will shed light on how people do learn and plied in the method of retrospection itself. In
might better learn from history. retrospect, we seem to perceive the logic of the
events which unfold themselves in a regular or
The two hypotheses explored are: (a) linear fashion according to a recognizable pattern
with an alleged inner necessity. So that we get
Preparation of this report was supported by the the impression that it really could not have hap-
Advanced Research Projects Agency of the De- pened otherwise, (p. 369)
partment of Defense (ARPA Order 2449) and
was monitored by the Office of Naval Research An apt name for this hypothesized tend-
under Contract No. N00014-73-C-0438 (NR 197- ency to perceive reported outcomes as hav-
026). The research reported constitutes part of ing been relatively inevitable might be
a doctoral dissertation submitted to The Hebrew "creeping determinism"—in contrast with
University of Jerusalem. I am deeply indebted to
Amos Tversky, Daniel Kahneman, Paul Slovic, philosophical determinism, which is the con-
Ruth Beyth, and Sarah Lichtenstein for their con- scious belief that whatever happens has to
tributions to this project. The detailed comments happen.
of two anonymous reviewers on a previous draft Phenomena resembling creeping determi-
are gratefully acknowledged. nism have been noted by psychologists as
Requests for reprints should be sent to Baruch
Fischhoff, Oregon Research Institute, P.O. Box well as historians. One example is Tversky
3196, Eugene, Oregon 97403. and Kahneman's (1971) "law of small
288
 HINDSIGHT *£ FORESIGHT 289

numbers," the belief that data which were possible outcomes of the event [including that
observed more or less had to be observed. A which actually happened—for After subjects] in
the light of the information appearing in the
second example is the tendency to rework passage.
or reconstruct the biographies of deviants
to show that their present diagnoses (labels) A typical passage, as taken from Woodward's
are inevitable products of their life histo- (1938) The Age of Reform, was:
ries (Lofland, 1969; Rosenhan, 1973; Schur, [1] For some years after the arrival of Hastings
1971). A third is the defensive attribution as governor-general of India, the consolidation
of responsibility for accidents, a process in of British power involved serious war, [2] The
which people carefully scrutinize the data first of these wars took place on the northern
frontier of Bengal where the British were faced
describing accidents in order to uncover or by the plundering raids of the Gurkas of Nepal.
impose a pattern that will increase their per- [3] Attempts had been made to stop the raids
ceived predictability and avoidability (Wals- by an exchange of lands, but the Gurkas would
ter, 1967). not give up their claims to country under British
control, [4] and Hastings decided to deal with
All of this evidence for creeping determi- them once and for all. [5] The campaign began
nism is, however, either indirect, imprecise, in November, 1814, It was not glorious. [6]
unsystematic (anecdotal), or confounded by The Gurkas were only some 12,000 strong; [7]
motivational and emotional issues. Experi- but they were brave fighters, fighting in terri-
tory well-suited to their raiding tactics. [8J
ment 1 directly tested the validity of the The older British commanders were used to war
creeping determinism hypothesis and ex- in the plains where the enemy ran away from
plored some of the concomitant effects of a resolute attack. [9] In the mountains of Nepal
outcome knowledge on judgment. it was not easy even to find the enemy. [10]
The troops and transport animals suffered from
the extremes of heat and cold, [11] and the offi-
EXPERIMENT 1 cers learned caution only after sharp reverses.
[12] Major-General Sir D. Octerlony was the
Method one commander to escape from these minor de-
feats, (pp. 383-384)
Design. The six subexperiments described in
this section are identical except for the stimuli The possible outcomes offered were: (a) Brit-
used. In each, subjects were randomly assigned ish victory, (b) Gurka victory, (c) military stale-
to one of five experimental groups, one Before mate with no peace settlement, and (d) military
group and four After groups. In each subexperi- stalemate with a peace settlement. For After sub-
ment, the Before group read a brief (150 word) jects, the appropriate outcome was appended to
description of a historical or clinical event for the passage in the form of an additional sentence,
which four possible outcomes were provided. The such as, "The two sides reached a military stale-
After groups read identical passages to which a mate, but were unable to come to a peace settle-
final sentence presenting one of the possible out- ment."
comes as the "true" outcome had been added. As Following the passage, subjects were asked, "In
the possible outcomes were mutually exclusive,
the light of the information appearing in the pas-
three of the four After groups received "true" sage, what was the probability of occurrence of
outcomes that actually had not happened. Sub-
each of the four possible outcomes listed below.
jects in all groups were asked to (a) estimate the
(The probabilities should sum to 100%)." On the
likelihood of occurrence of each of the four pos- following page, each datum appeared on a separate
sible outcomes, and (b) evaluate the relevance of line followed by a seven-point scale on which
each datum in the event description. In two of subjects were asked to indicate "how relevant or
the subexperiments subjects were also asked to important each datum in the event description was
indicate the relative extent to which they relied on in determining the event's outcome." The numbers
the passage and on outside information. in the passage above indicate the division into
datum units. They did not appear in the passage
Instructions. The cover sheet of each question- presented to subjects.
naire read: Stimulus selection. Four different events were
used to achieve greater generality for the results
In this questionnaire we are interested in obtained: Event A, the British-Gurka struggle
knowing how people judge the likelihood of cited above; Event B, the near-riot in Atlanta,
possible outcomes of social events. A passage Georgia in July 1967, as described in the Kerner
describing an unfamiliar historical event ap- Commission Report on Civil Disorders (National
pears below. We will ask you to evaluate the Advisory Commission, 1968, pp. 53-5); Events
probability of occurrence of each of the four C and D, clinical cases reported by Albert Ellis
290 BARUCH FISCHHOFF

(1966, pp. 262-264).* For Events C and D, the edge of statistics). For the sake of tabular
word "social" in the instructions was replaced by brevity, only one subexperiment in each pair
"individual" and the word "historical" was de-
leted. is presented.
Several methodological considerations guided the The creeping determinism hypothesis pre-
event selection process: (a) The event should be dicts that After subjects told that a particu-
sufficiently familiar to permit intelligent responses lar outcome has happened will assign it a
and sufficiently unfamiliar to rule out the possibility
of subjects knowing what really happened—espe- higher probability than will Before subjects.
cially those receiving false outcome reports, (b) Four outcomes reported to different groups
Past events were used to allow provision of "true" in each of six subexperiments afford 24 op-
outcomes to the After groups.2 (c) The space of portunities to test the hypothesis. The criti-
possible outcomes had to be readily partitionable.
For Events B, C, and D, the set of outcomes was cal comparisons are between the outlined
constructed to be mutually exclusive and exhaus- diagonal cells (those indicating the mean
tive. Although this is not the case for Event A, probability assigned to an outcome by sub-
pretests indicated that these four outcomes con- jects for whom that outcome was reported to
stituted an effective partition.
Subjects. Approximately equal numbers of sub- have happened) and the Before cell in the
jects participated in each group in each subexperi- top row above them.
ment. Event A (Gurkas) was administered twice, In each of the 24 cases, reporting an out-
once to a group of 100 English-speaking students come increased its perceived likelihood of
recruited individually at The Hebrew University occurrence (p < .001; sign test). Twenty-
campus in Jerusalem and once to a class of 80
Hebrew-speaking subjects at the University of the two of these differences were individually
Negev in Beer Sheba. Event B (riot) was ad- significant (p < .025; median test). Thus
ministered to two separate classes at The Hebrew the creeping determinism effect was obtained
University, one containing 87 Hebrew-speaking
psychology majors with at least one year's study of over all variations of subject population,
statistics and one of 100 Hebrew-speaking stu- event description, outcome reported, and
dents with no knowledge of statistics. Event C truth of outcome reported. The differences
(Mrs. Dewar) was administered to the 80 Uni- between mean Before and After probabili-
versity of the Negev students; Event D (George) ties for reported outcomes ranged from
to the 100 Hebrew University students without
statistics training. 3.6% to 23.4%, with a mean of 10.8%.
Procedure. Questionnaires for the various ex- Slightly over 70% of After subjects assigned
perimental groups were distributed randomly. the reported outcome a higher probability
Subjects devoted 20-30 min to the completion of than the mean assignment by the corre-
each questionnaire.
sponding Before subjects.
Results No outcome was judged inevitable by any
Probability estimates. Table 1 presents Before subject, whereas a small proportion
the mean probability assigned to each out- (2.1%) of After subjects did assign 100%
come by subjects in each experimental group to reported outcomes. Evidently, most After
for each subexperiment. subjects felt that in the light of the facts
Similar patterns of data emerged in the given in the description, other (unreported)
two subexperiments using Event A (differ- outcomes were still possible (e.g., "The
ing in subjects' language) and in the two Gurkas had a 70% chance of winning, but
using Event B (differing in subjects' knowl- the British still might have pulled it off").
Similarly, After subjects found a higher per-
1
Copies of these stimuli with the four offered centage of unreported outcomes to have been
outcomes will be supplied with all requested re- impossible (as indicated by a probability of
prints. The permission granted by the Oxford
University Press and Ronald Press to use these 0%) than did Before subjects (11.5% ver-
copyrighted materials is gratefully acknowledged. sus 8.0%).
2
It might be wondered whether Before subjects
might not behave like After subjects with reference Another way to appraise the extent of
to their predictions, which they know are actually creeping determinism is to translate mean
postdictions. In a series of five experiments Before probabilities into the form of a priori
(Fischhoff, in press), we found that manipulating
the temporal setting of possible outcomes has no odds and the mean After probabilities for
effect on their perceived likelihood. reported outcomes into a posteriori odds.
 HINDSIGHT ^ FORESIGHT 291

TABLE 1
MEAN PROBABILITIES ASSIGNED TO EACH OUTCOME, EXPERIMENT 1
Outcome evaluated
Outcome
Experimental group provided

Event A: British-Gurka struggle (English-speaking subjects)

Before 20 none 33.8 21.3 32.3 12.3

After 20 1 57. 2 14.3 1S.3 13.4
20 2 30.3 38,4 20.4 10.5
20 3 25.7 17.0 48.0 9.9
20 4 33.0 15.8 24.3 27.0

Event B: Near-riot in Atlanta (subjects with knowledge of statistics)

Before 20 none 11.2 30.8 43.8 14.2

After 20 1 [ 30.6 25.8 23.3 20.3
20 2 5.5 51.8 24.3 18.5
20 3 3.9 23.9 50.8 21.4
20 4 16.7 31.9 23.4 27.9

Event C: Mrs. Dewar in therapy

Before 19 none 26 6 15.8 23.4 34.4

After 13 1 43 1 13.9 17.3 25.8
17 2 26 5 23.2 13,4 36.9
16 3 30 6 14.1 34.1 21.3
17 4 21 2 10.2 22.6 46.1

Event D: George in therapy

Before 17 none 27.4 26.9 39.4 6.3

After 18 1 33.6 20.8 37.8 8.0
18 2 22.4 41.8 28.9 7.1
20 3 20.5 22.3 50.0 7.3
17 4 30.6 19.5 37.7 12.3

Note, The actual outcomes are numbers 1, 1 , 4 , and 2 for Kvents A, B, C, and D, respectively. Outlined cells are those with
After probabilities of reported outcomes.

The ratio of prior and posterior odds for subjects' expectations to varying degrees.
outcome i provides a sort of average likeli- There was a highly significant negative cor-
hood ratio for the impact of the datum relation ( r = . 4 3 5 ; p < .001) between the
"Outcome i did actually occur" (where the prior odds and likelihood ratios associated
two hypotheses are "Outcome i occurs" and with reported outcomes (as computed in the
"Outcome i does not occur"). Over the preceding paragraph). Thus, the more un-
24 outcomes reported, these likelihood ratios likely an outcome report, the greater the
varied from 1.2 to 3.5 (M = 1.96). Thus impact it has.
in the" present sense, reporting an outcome's Relevance judgments. Table 2 presents
occurrence approximately doubles its per- the mean relevance judgments for each
ceived likelihood of occurrence. datum in one subexperiment. Inspection re-
Because the outcomes varied considerably veals that the relevance attributed to any
in their mean Before probability (from 6.3% datum is highly dependent on which out-
to 44.0%), reporting their occurrence may come, if any, subjects believe to be true.
be seen as confirming (or disconfirming) Some of these differences seem readily inter-
292 BARUCH FISCHHOFF

TABLE 2
MEAN DATA RELEVANCE JUDGMENTS FOR EVENT A,
EXPERIMENT 1 (HEBREW-SPEAKING SUBJECTS)

Datum number
Outcome reported l 2 3 4 5 6 7 8 9 10 11 12
None 4.50 5.11 4.22 5.78 4.50 6.00 5.50 5.44 4.39 4.56 4.28 5.56
British victory 4.78 4.44 5.28 4.83 4.61 4.44 4.61 4.56 5.72 5.33 5.78 4.11
Gurka victory 3.66 4.83 3.55 4.44 5.89 5.11 4.11 4.61 3.72 5.22 4.11 4.78
Stalemate
with peace treaty 4.50 4.72 4.55 5.89 5.50 4.17 4.22 5.00 4.22 5.22 4.89 4.94
without peace treaty 4.94 5.50 4.39 5.11 5.33 5.11 4.78 4.39 4.17 3.72 4.50 4.61

pretable. For example, the fact that "the having greater mean relevance for one out-
British officers learned caution only after come than another, (c) Datum-evaluated
sharp reverses" (Datum No. 11) was effects appear in all but one subexperiment.
judged most relevant by subjects told of a They reflect data perceived to be either rele-
British victory, and rather irrelevant by sub- vant or irrelevant whatever happens (e.g.,
jects told of a Gurka victory. Hastings' decision to deal with the Gurkas
A less impressionistic analysis on the ef- "once and for all" [Datum No. 4] was uni-
fects of outcome knowledge on relevance versally perceived as relevant).
judgments proceeded in the following man- Reliance. Subjects in two of the sub-
ner. For each subexperiment, a two-way experiments were asked to indicate with a
(outcome reported X datum evaluated) fixed number from Q% to 100% the extent to
model analysis of variance (ANOVA) was which they had relied on the material pre-
performed on subjects' judgments of data sented in the passage compared with general
relevance. To accommodate the varying (outside) knowledge. In both cases, Before
number of subjects in the experimental subjects indicated relying significantly (/> <
groups of subexperiments, the following pro- .05; median test) more on the passage than
cedure was adopted: The analysis was re- did After subjects.
peated three times to produce maximum,
minimum, and middle solutions. For the Discussion
maximum solution, subjects were randomly Reporting an outcome's occurrence con-
sampled from the smaller experimental sistently increases its perceived likelihood
groups and their responses duplicated, equat- and alters the judged relevance of data de-
ing the size of all cells. For the minimum scribing the situation preceding the event.
solution, subjects were randomly deleted Hindsight-foresight differences in perceived
from the larger groups until cell size was data relevance have also been noted by his-
equated. For the middle solution, a com- torians observing the creeping determinism
bination of duplication and elimination was effect. Consider, for example, Tawney
performed. (1961): "Historians give an appearance of
The same ANOVA was performed on the inevitability to an existing order by dragging
three sets of data. The results discussed into prominence the forces which have tri-
hold for all three solutions, (a) In each umphed and thrusting into the background
subexperiment, there was a significant Out- those which they have swallowed up" (p.
come Reported X Datum Evaluated interac- 177). Or, consider Wohlstetter (1962):
tion reflecting the differential effect of out- "It is much easier after the event to sort
come knowledge on relevance judgments the relevant from the irrelevant signals.
discussed above and shown in Table 2. (b) After the event, of course, a signal is always
Over the six subexperiments, only two weak crystal clear. We can now see what disaster
outcome-reported effects emerged. Thus, it was signaling since the disaster has oc-
there is no indication of an entire set of data curred, but before the event it is obscure and
 HINDSIGHT 76 FORESIGHT 293

pregnant with conflicting meanings" (p. edge has on their perceptions was examined
387). in Experiment 2. Aside from helping to
How justified are the judgmental changes clarify the nature of creeping determinism,
affected by receipt of outcome knowledge? these results have considerable intrinsic in-
It is hard to say, simply because there is no terest. Awareness is clearly crucial to know-
unequivocal way to assign probabilities or ing what one has learned from the past
relevance weights to unique events like the (i.e., from outcome knowledge). It may be
British-Gurka struggle. If, for example, necessary for learning from the past at all.
someone claims that there was no chance
(or a 7% chance or a 98.6% chance) of a EXPERIMENT 2
thermonuclear war during the 1960s, who Method
can prove him wrong? Indeed, the only Design. Subjects were presented stimulus ma-
wrong estimate is that it was 100% likely. terials identical to those used in the After groups
Were such events well defined and re- of Experiment 1, with each event description ac-
companied by a "true" outcome. They were asked
occurring, the wisdom of increasing the to respond "as they would have had they not
postdicted probability of some reported out- known the outcome." For each of the four events
comes would be readily apparent. Consider there were four After (ignore) groups, one re-
a judge who has previously encountered four ceiving each possible outcome as true. If sub-
jects are aware of the effect of outcome knowl-
British-Gurka-type struggles, each indistin- edge on their judgments, the responses of all of
guishable from the one used here, two of the After (ignore) groups should resemble those
which were won by the Gurkas. Upon of that Before group in Experiment 1 which dealt
learning of another Gurka victory, he may with the same event. If After (ignore) subjects
are completely unable to ignore the effect of out-
properly update that outcome's predictive come knowledge, their responses should resemble
(Before) probability of 50% to a higher those of the After group in Experiment 1 which
postdictive (After) probability. Hearing of received the same outcome as "true."
a Gurka victory may also justify some Instructions. The cover of each test booklet
read:
change in relevance judgments by showing,
for example, the true importance of British A number of short descriptions of real social
suffering from climatic extremes. It may and personal events appear below, each with a
number of possible outcomes. On the basis of
also teach him something about the nature these data, we ask you to evaluate the likelihood
of 19th century colonialism—and thus change of the outcomes listed. We thank you for your
the sort of "laws" or reasons he uses in participation.
drawing inferences from the event descrip- Each remaining page of the test booklet was identi-
tion. cal to the corresponding page of the Experiment 1
Thus, the judgmental changes which we booklet, except that each response section was
have called creeping determinism could con- preceded with the instruction to "answer as you
would have had you not known what happened."
ceivably reflect what judges learn from out- Subjects. Eighty members of an introductory
come reports. The skepticism expressed in statistics class at the University of the Negev
the anecdotal observations presented above participated.
(e.g., Florovsky, Schur, Tawney), however, Procedure. Questionnaires were randomly dis-
tributed to a single group of subjects. Each sub-
suggests that this is not the case. In the ject received one version of each of the four dif-
light of these comments, it appears that vvhat ferent events. In a test booklet, Events A, B, and
passes for the wisdom of hindsight often C alternated systematically as the first three events,
contains heady doses of sophistry—that the with Event D (the least interesting) always ap-
pearing last. Order was varied to reduce the
perceived inevitability of reported outcomes chances that subjects sitting in adjoining seats
is imposed upon, rather than legitimately in- either copied from one another or discovered the
ferred from, the available evidence. experimental deception. All materials were in
As described in these accounts, postdictive Hebrew. Questionnaires were anonymous.
likelihood estimates are exaggerated through
Results
a largely unconscious process evoked by re-
ceipt of outcome knowledge. How aware Probability estimates. Table 3 presents
people are of the effect that outcome knowl- mean probability assignments by subjects in
294 BARUCH FISCHHOFF

TABLE 3
MEAN PROBABILITIES ASSIGNED BY SUBJECTS RESPONDING "As IF You DID NOT
KNOW WHAT HAPPENED," EXPERIMENT 2
Outcome evaluated
Outcome
Kxperimental group provided

Event A: British-Gurka struggle

Before 17 none 29.4 23.S 34.7 12.4

After (ignore) 20 1 129.8 27.4 24.9 18.4
IS 2 38.0 21.7 19.7 20.7
18 3 22.1 31.8 31.9 14.3
18 4 18.1 32.9 28.9 21.2

Event B: Near-riot in Atlanta

Before 39 none 11.3 29.0 43.9 16.3

After (ignore) 17 1 [24.6 27.0 28.3 19.8
21 2 9.0 41.5 36.4 13.1
20 3 6.3 24.5 43.5 25.8
20 4 13.3 20.3 36.5 24.0

Event C: Mrs. Dewar in therapy

Before 19 none 26.6 15.8 23.4 34.4

After (ignore) 19 1 36.4 10.2 16.1 37.4
19 2 24.7 28.8 15.5 31.9
15 3 25.1 13.7 34.9 26.4
20 4 18.3 12.3 21,8 52.8

Event D: George in therapy

Before 17 none 26 4 26.9 39.4 6.3

After (ignore) 17 1 41 8 16.5 35.3 6.5
18 2 24 6 35.9 32.4 7.0
20 3 18 3 20.4 57.3 4.0
18 4 21 0 21.1 38.4 19.6

Note. In each case the Before results are taken from the corresponding before (no outcome) group in Experiment 1 (subjects
who actually responded not knowing what happened). Outlined cells are those with After(ignore) probabilities of repo rte
outcomes.

each of the After (ignore) groups along with the reported outcome was higher than the
the responses of the corresponding Before mean Before probability for the same event.
groups from Experiment 1. (The Hebrew- For reported outcomes the mean Before-
speaking group is used for Event A, the After (ignore) difference of 9.2% was
pooled responses of both relevant subexperi- slightly but not significantly less than the
ments for Event B.) The entries in each 10.8% mean Before-After difference in Ex-
row will be called a profile. They indicate periment 1 (p > .10; Mann-Whitney U
the probabilities subjects believed they test).
would have assigned to the outcomes had The After (ignore) profiles closely resem-
they not known "what really happened." bled the corresponding After profiles. For
These reconstructed probabilities indicate 14 of 16 profiles, the mean absolute differ-
no more than marginal awareness of the ef- ence between corresponding cells was smaller
fects of outcome knowledge. In 13 of 16 for the After(ignore)-After comparison than
cases, the mean After (ignore) probability of for the relevant After (ignore)-Before com-
 HINDSIGHT ^ FORESIGHT 295

parison (/> < .002; sign test). The median in the direction predicted by the creeping
absolute difference between corresponding determinism hypothesis. Experiment 2 has
cells was 3.7% for After (ignore)-After, shown that subjects are either unaware of
and 6.4% for After (ignore)-Before (p < outcome knowledge having an effect on their
.001; Mann-Whitney U test). There is no perceptions or, if aware, they are unable to
apparent reason, other than sampling error, ignore or rescind that effect. Both the
for the weaker results obtained with Event A. relevance and the probability judgments of
Relevance judgments. If After(ignore) After (ignore) subjects suggest that subjects
subjects are able to ignore outcome knowl- fail to properly reconstruct foresightful (Be-
edge, the outcome report they received fore) judgments because they are "an-
should have no effect on their reconstructed chored" in the hindsightful state of mind
relevance judgments. Instead, however, created by receipt of outcome knowledge.
these relevance judgments clearly reflected It might be asked whether this failure to
the outcomes that After (ignore) subjects empathize with ourselves in a more ignorant
believed to have happened (but were in- state is not paralleled by a failure to empa-
structed to ignore). For example, in Ex- thize with outcome-ignorant others. How
periment 1 After subjects told of a British well people manage to reconstruct the per-
victory assigned substantially greater impor- ceptions that others had before the occur-
tance to the fact that "British officers learned rence of some event is a crucial question for
caution only after sharp reverses" (Datum historians, and indeed for all human under-
No. 11) than did Before subjects; those told standing. The assumption that we clearly
of a Gurka victory assigned it slightly less perceive how others viewed situations be-
importance. After (ignore) subjects in Ex- fore receipt of outcome knowledge underlies
periment 2 who were asked to ignore a most second-guessing of their decisions. Ex-
report of British victory believed that even periment 3 examined this question.
without the report they would have per-
ceived the relevance of Datum No. 11; those EXPERIMENT 3
told to ignore a report of Gurka victory be- Method
lieved that they in foresight would have
seen its irrelevance. When the relevance Design. Subjects were presented with stimulus
judgment ANOVA of Experiment 1 is re- materials identical to those used in Experiments
1 and 2. They were asked to respond as had other
peated on the present data, this dependence student judges who had not known the true out-
is reflected in highly significant (p < .0005) come. Before (others) subjects were not provided
Outcome Reported X Datum Evaluated in- with any outcome knowledge. After (others) sub-
teractions. jects received versions of the stimulus events with
one of the four possible outcomes presented as
Interestingly, for 128 of the 184 individual the true outcome (what had actually happened).
datum units evaluated by subjects in the After(others) subjects' task was essentially to ig-
four outcome groups of the four events After nore outcome knowledge in order to respond like
and After(ignore) relevance judgments were Before (others) subjects.
Instructions. The cover of each test booklet
either both higher or both lower than the read:
corresponding Before judgments (as was the
case in the example, Datum No. 11, given Short descriptions of a number of real social and
personal events appear below, each with several
above) (#=5.23; sign test). There was possible outcomes. These descriptions were pre-
no tendency for After and After (ignore) sented to students of social science in other uni T
relevance judgments to be consistently versities in Israel. (However, they were not
higher or lower than Before relevance judg- told which of the possible outcomes actually
happened.) We will ask you to guess the judg-
ments, which might in itself account for ments of these students regarding the likeli-
this result. hood of possible outcomes. We thank you for
your participation.
Discussion
The section in parentheses only appeared in the in-
Experiment 1 showed that receipt of out- structions for After (others) subjects. Each page
come knowledge affects subjects' judgments of the test booklets was identical to the corre-
296 BARUCH FISCHHOFF

TABLE 4
MEAN PROBABILITIES ASSIGNED BY SUBJECTS RESPONDING "As DID OTHER STUDENTS
WHO DID NOT KNOW WHAT HAPPENED," EXPERIMENT 3

Outcome evaluated
Outcome
Experimental group provided

Event A: British-Gurka struggle

Before 21 none 26.4 24.S 29.5 19.5

After (ignore) 17 1 39.4 22.4 20.3 18.8
17 2 18.8 42.6 20.3 20.0
22 3 31.1 21.2 26.6 20.0
17 4 28.2 21.9 23.7 26.2

Event B: Near-riot in Atlanta

Before 20 none 11.0 24.0 41.8 23.2

After (ignore) 17 1 15.0 24.7 36.5 23.8
18 2 13.2 36.0 35.2 14.6
19 3 4.8 22.5 51.1 21.6
16 4 12.3 26.4 38.4 22.8

Event C: Mrs. Dewar in therapy

Before 21 none 19.6 15.9 24.0 40.5

After (ignore) 18 1 20.3 20.0 28.3 31.4
18 2 31.9 23.3 14.8 30.0
16 3 30.6 12.5 26.9 30.1
19 4 12.5 20.4 22.6 44.4

Event D: George in therapy

Before 19 none 30.7 22.4 39.2 7.8

After (ignore) 15 1 46.0 15.3 30.0 8.7
16 2 22.5 36.6 34.1 6.9
17 3 19.8 14.8 57.7 7.8
16 4 23.5 18.3 40.3 17.8
Note, Outlined cells are those with After (ignore) probabilities of reported outcomes.

spending page of the Experiment I test booklets, come-ignorant others than did Before (oth-
except for the addition of a reminder, "Answer as ers) subjects. As in Experiment 2, being
you think other students (who did not know what
happened) answered" before each response section. told to ignore outcome knowledge slightly,
Subjects, Ninety-four members of an inter- but not significantly (p > .10; Mann-Whit-
mediate statistics class at the University of the ney U test), reduced its impact. The mean
Negev participated. Before(others)-After(others) difference was
8.7% compared with the mean Before-After
Results difference of 10.8% in Experiment 1.
Probability estimates. Table 4 presents Relevance judgments. After (ignore) sub-
mean probability assignments by subjects in jects who had received different outcome re-
each group. After (others) subjects' in- ports attributed markedly different relevance
ability to ignore the effects of creeping de- judgments to the outcome-ignorant others.
terminism is clearly evident. For 14 of the The dependence of the relevance judgments
16 reported outcomes (p < .002; sign test), that they attributed on the outcome knowl-
they attributed higher probabilities to out- edge they were to ignore produced signifi-
 HINDSIGHT ^ FORESIGHT 297

cant (p < .01) Outcome Reported X Datum of the visits (e.g., Nixon's meeting Chair-
Evaluated interactions for each of the four man Mao, visiting Lenin's tomb, or an-
events. Thus, After (ignore) subjects ex- nouncing that the trips were successful).
pected other subjects to have seen in fore- From 2 wk to 6 mo after the trips' comple-
sight patterns of data relevance that they tion, these same subjects were asked to re-
themselves only saw in hindsight. member as best they could their own origi-
Projection. Comparing Tables 1 and 4 nal predictions. They were also asked to
and Tables 3 and 4, it is apparent that the indicate for each event whether or not they
entries in corresponding Before and Before- believed that it had actually happened.
(others) cells are quite similar, as are cor- Results showed that subjects remembered
responding After (others) and After(ignore) having given higher probabilities than they
cells. The mean absolute difference between actually had to events believed to have oc-
entries in corresponding cells is 3.5% for the curred and lower probabilities to events that
first comparison, 5.1% for the latter. This hadn't occurred. Their original predictions
suggests that when asked to respond like showed considerable overestimation of low
similar others, subjects respond as they be- probabilities, that is, too many events that
lieve they themselves would have responded they judged to be extremely unlikely or im-
in similar circumstances (i.e., by projec- possible did occur. The probability judg-
tion). Both the probability and relevance ments that they remembered, however, con-
judgments of After (others) subjects more sistently underestimated low probabilities.
closely resembled those of After (ignore) Indeed, almost no events to which they re-
and After subjects than those of Before(oth- membered assigning low probabilities were
ers) subjects. perceived to have occurred.
Reasons. Some 87% of the subjects pro- Thus, undiagnosed creeping determinism
vided reasons for their judgments. Al- not only biases people's impressions of what
though content analysis of these reasons they would have known without outcome
proved intractable, one interesting finding is knowledge, but also their impressions of
that After (others) subjects offered consist- what they themselves and others actually did
ently more reasons than Before subjects (p know in foresight.
<.05; median test). In Experiment 1,
After subjects reported relying more on out- Explanations
side information (as compared with the text)
than did Before subjects. Perhaps in both The simplest hypothesis regarding the
cases knowing what happened facilitates manner in which judges process outcome
knowing where to look for and what to ac- knowledge suffices to account for these re-
cept as reasons. sults. Assume that upon receipt of outcome
knowledge judges immediately assimilate it
GENERAL DISCUSSION with what they already know about the event
in question. In other words, the retrospec-
Finding out that an outcome has occurred tive judge attempts to make sense, or a co-
increases its perceived likelihood. Judges are, herent whole, out of all that he knows about
however, unaware of the effect that outcome the event. The changes in relevance judg-
knowledge has on their perceptions. Thus, ments could reflect such assimilative mean-
judges tend to believe that this relative in- ing adjustment.
evitability was largely apparent in foresight, Assimilation of this type would tend to
without the benefit of knowing what hap- induce creeping determinism for judges
pened. using any of the techniques for producing
In a fourth study (Fischhoff & Beyth, subjective probability estimates appearing in
1975), subjects were asked on the eve of Tversky and Kahneman's (1974) com-
former President Nixon's trips to China pendium. Judges using the heuristic of
and the USSR (in early 1972) to estimate "representativeness" perceive outcomes as
the probability of various possible outcomes likely when they match or represent the
298 BARUCH FISCHHOFF

dominant features of the situation that tion, were unable to adequately unanchor
produced them. Assimilation of outcome themselves from the perspective of hind-
knowledge should certainly increase the per- sight.
ceived "fit" between the reported outcome
and the situation that preceded it. A second Implications
heuristic leads judges to evaluate an out- In the short run, failure to ignore out-
come's likelihood by the relative "avail- come knowledge holds substantial benefits.
ability" of scenarios leading to its occur- It is quite flattering to believe, or lead others
rence and nonoccurrence. The judge who to believe, that we would have known all
knows what happened, and has adjusted his along what we could only know with out-
perceptions in the light of that knowledge, come knowledge, that is, that we possess
may well find it difficult to imagine how hindsightful foresight. In the long run, how-
things could have turned out otherwise. ever, unperceived creeping determinism can
An alternative mode of explanation focuses seriously impair our ability to judge the
on structural differences between the tasks past or learn from it.
of hindsight and foresight. Judges possessing Consider a decision maker who has been
outcome knowledge may, for example, tend caught unprepared by some turn of events
to reverse their temporal perspective and and who tries to see where he went wrong
produce scenarios that proceed backward by recreating his preoutcome knowledge
in time, from the outcome to the preceding state of mind. If, in retrospect, the event
situation. Such scenario retrodiction may appears to have seemed relatively likely, he
effectively obscure the ways in which events can do little more than berate himself for
might not have taken place, much as solving not taking the action which his knowledge
a maze backward can obscure the ways in seems to have dictated. He might be said
which one might have gotten lost entering to add the insult of regret to the injury
from the beginning. inflicted by the event itself. When second-
Receipt of outcome knowledge may also guessed by a hindsightful observer, his mis-
restructure the task of judges using the fortune appears to have been incompetence,
"anchoring and adjustment" heuristic (Tver- folly, or worse.
sky & Kahneman, 1974). Judges may esti- In situations where information is limited,
mate the likelihood of a reported outcome by and indeterminate, occasional surprises-—•
initially assigning it 100%, the most salient and resulting failures—are inevitable. It
possible value, and then looking for reasons is both unfair and self-defeating to castigate
to adjust downward from there. Adjust- decision makers who have erred in fallible
ment from initial values is typically inade- systems, without admitting to that fallibility
quate and would produce creeping determ- and doing something to improve the system.
inism with this task. According to historian Roberta Wohlstetter
Of these explanations, those based on (1962), the lesson to be learned from Amer-
assimilation most readily account for the ican surprise at Pearl Harbor is that we
underestimation of creeping determinism must "accept the fact of uncertainty and
found in Experiments 2 and 3 and in the learn to live with it. Since no magic will
Fischhoff and Beyth (1975) study. Making provide certainty, our plans must work
sense out of what one is told about the past without it" (p. 401).
seems so natural and effortless a response When we attempt to understand past
that one may be unaware that outcome events, we implicitly test the hypotheses or
knowledge has had any effect at all on him. rules we use to both interpret and anticipate
Judges who are aware that outcome knowl- the world around us. If, in hindsight, we
edge has affected their perceptions still face systematically underestimate the surprises
the unenviable task of reconstructing their which the past held and holds for us, we are
foresightful state of mind. "Undiagnosed subjecting those hypotheses to inordinately
creeping determinism" would characterize weak tests and, presumably, finding little
the responses of subjects who, in reconstruc- reason to change them. Thus, the very out-
 HINDSIGHT 96 FORESIGHT 299

come knowledge which gives us the feeling things. Organisational Behavior and Human
that we understand what the past was all Performance, 1975, 13, 1-16.
about may prevent us from learning anything Florovsky, G. The study of the past. In R. H.
Nash (Ed.), Ideas of history (Vol. 2). New
3
from it. York: Button, 1969.
Lofland, R. Deviance and identity. Englewood
3 Cliffs, N.J.: Prentice-Hall, 1969.
Elaboration on this point as well as specula- National Advisory Commission. Kerner Com-
tion on how hindsight can be improved may be
mission report on civil disorders. New York:
found'in Fischhoff (1974) and Fischhoff and Beyth Bantam, 1968.
(1974). Rosenhan, D. On being sane in insane places.
Science, 1973, 79, 250-252.
REFERENCES Schur, E. Labelling deviant behavior. New York:
Harper & Row, 1971.
Ellis, A. Psychosexual and marital problems. In Tawney, R. H. The agrarian problem in the
L. A. Berg & L. A. Pennington (Eds.), An sixteenth century. New York: Franklin, 1961.
introduction to clinical psychology. New York: Tversky, A., & Kahneman, D. Belief in the law
Ronald, 1966. of small numbers. Psychological Bulletin, 1971,
Fischhoff, B. Hindsight: Thinking backward? 76, 105-110.
Oregon Research Institute Research Monograph, Tversky, A., & Kahneman, D. Judgment under
1974,14, No. 1. uncertainty: Heuristics and biases. Science, 1974,
Fischhoff, B. Temporal setting and judgment 185, 1124-1131.
under uncertainty. Organisational Behavior Walster, E. "Second guessing" important events.
and Human Performance, in press. Human Relations, 1967, 20, 239-250.
Fischhoff/ B., & Beyth, R, Failure has many Wohlstetter, R. Pearl Harbor: Warning and
fathers (Review of Victims of groupthlnk by decision. Stanford, Calif.: Stanford University
I. Janis). Oregon Research Institute Research Press, 1962.
Bulletin, 1974, 14, No. 17. Woodward, E. L. Age of reform. London:
Fischhoff, B., & Beyth, R. "I knew it would hap- Oxford University Press, 1938.
pen"—remembered probabilities of once-future (Received January 6, 1975)