WWWW ' ' I ' IIII IM UMiiii<iijpji|iyipii«|JHUi--.iJuujji iPi|iwi,wu«i>«,p,yi|, jiiiuipuii4iii|iet!<lipi|ip.Wiiy wupy ■KIIU.WiHMW IMI

AD-A008 591

TEMPORAL SETTING AND JUDGMENT UNDER UNCERTAINTY

Baruch Fischhoff

Oregon Research Institute

.

P' spared for:

Office of Naval Research Advanced Research Projects Agency

7 April 1975

DISTRIBUTED BY:

mi] National Technical Information Service U. S. DEPARTMENT OF COMMERCE

- -'- -—'■ -■- --^-•^»^^^^■^^•-^^-■^^•■'■■--•■- -■ ■ ■■ -.-^.^-..... - .........

Hin JjJIipiilllipiiMPiP mmmi*mmmmmmmmmmmmmmm

ONR Technical Report

125127

TEMPORAL SETTING AND JUDGMENT UNDER UNCERTAINTY

Baruch Fischhoff

D D C

mmmum B .

This research was supported by the Advanced H^mch Projects Agency of thr 8fenso (ARPA Order No. 2^9) ang ^onitired by^NR unde^

Contract No^jg fej3-C-04.td (MR 197-026}.

nt are those of the ssarily represent;« | the the Advanced Research r the U.S. Government

ny purpose of the ites Government.

Research Bulletin Vol. 14, No. K.

^mmmunrnm m

-J»WJP*MAfiW«^MPIi|up|(p^ (.pWWMWWflW'^U^'l'^J^'-t-LUlilkWIlM^I'JUilllÄJ^^WlIU

MM

Vn^ll^gg<f^,^f,

Security CUitifiotiop

DOCUMENT CONTROL DATA -R&D tSteutllr cl„,Hlc»'lon ol IMI«. body ot mbHtmct mnd Indtwlnj »nnol.llon mutt 6. .nf f tf wh»n Ih* ovrmll fporl I« cffiUtdj

. . .„ ,- Si gSS! lla. RKPORT ■KCUntTV CL*»iiriC»TION OmaiNATINO ACTIVITY fCorpor»«» «Millar;

Oregon Research Institute Eugene, Oregon

Unclassified 2h. GROUP

3 RCPOUT TITLE

Temporal Setting and Judgment Under Uncertainty

4. DCSCniPTi VC NOT«» (Typ* ol topotl ond In

Technical Report S. «uTHORItl (Flnl nmir; mlddl* Inlllml, Imtt nmato)

Baruch Fischhoff

• REPORT DATE

April 7, 1975

■Ta. TOTAL NO. OP PASEI

f. CONTRACT OR GRANT NO.

NOOOU-7'J-C-OA33 b. PROJECT NO.

NR (197-026) C.

Arpa Order No. 2A49

?J 7b. NO. OP REPS

22 M. ORIGINATOR'» REPORT NUMBCRI»»

Oregon Research Institute Research Technical Report Vol. 14, No. 14.

9b. OTHER REPORT NO(»l (Anr olhmr numbon Out may b« —l0t*J (Ml nporl)

10. DISTRIBUTION STATEMENT

This document has been approved for public release and sale. Its distribution

is unlimited

II. SUPPLEMENTARY NOTEJ 12. SPONSORING MILITARY ACTIVITY

Office of Naval Research Code 455 Arlington, Virginia 22217

13 ABSTRACT

In a series of experiments, subjects judged the likelihood of events set either in the past or the future. No consistent differences were found in either the central tendency or the dispersion of subjects' likelihood judgments regarding past and future events which differed solely in their temporal setting. Temporal setting was, however, found to affect the pro- duction of possible event outcomes. These results contradict a "sure past" hypothesis, which has been advanced by a number of observers according to which judges are more confident in dealing with past than future events. They also eliminate a possible source of methodological difficulty in existing judgmental studies and provide some insight into the use of judg- mental heuristics. Belief in the "sure past" hypothesis is discussed as the result of confusion between temporal setting and its ecological correlates.

*J. DD .^..1^73 REPLACE* DO PORM «47«. I J*H •«. WHICH I«

OBSOLETE POR ARMY USE. «MUM fflftH6E

HiajiiMaMiiai l*^^mim*mmii*mttmiimmamdmiii^mm WU^. ^uäiMtit^äUMMiamämiummikst ~ "■-j- ' ■•

^JMJUWMjlIM .^. W^

Uncla88lfled

K«V WOROt

JUDGMENT

PREDICTION

DECISION MAKING

INFORMATION PROCESSING

UNCERTAINTY

■■.■ i J •'■ ; ■ 1 •

II

LINK A

HOL«

LtNK m

NOLI I WT

LINK C

NQLB

m?

Sactmty Clanslflcattoti

"-"—-—"'-- ^~*^~*^-'~^.*. ^■■■^..■-■■^..-.■■^.s^^^jM^^^ai^^^ ..„.■...„,.„:,. ■.„..:,„...^.,J..^. ^.w.t-^^^^.-aija.^MMMAA»^ Mil

U _-,^llBUil|PlppiiBP«J^

OREGON RESEARCH INSTITUTE

TEMPORAL SETTING AND JUDGMENT

UNDER UNCERTAINTY

Baruch Fischhoff

Oregon Research Institute

[

Oregon Research Institute Research Bulletin

Vol. 14 Number 14

1974

-—— ..,^^-,.,-^>. ....^^aaai^^^^v

■ i ^ -wiip.upyitpwipiprjiJiHipJ.,,,Jjii4J.i41 '3rwBnwssww5^l9(P!P^^iiB|(ppip(PpipB^W||BB^

Temporal Setting

2

Temporal Setting and Judgment Uider Uncertainty

Are likelihood Judgments affected by :he temporal setting of

the events being judged; i.e.. by whether those events are set in the past

or future. Aside from its obvious suUtantive interest for the psychology

of time and the study of judg.ent, the answer to this question has methodo-

logical implications for both fields.

Research In Judgment hns typically Ignored the temporal aettlng of the

stlMuli uaed in experimental taaks. A suhstantlal temporal netting effect

„ould »eggest a need to review existing findings with sn eye to the setting

of the stimuli used, and at the extreme, the develcpmant of diatinet pay

chologles of predictive and poetdlctive Judgment. Similarly, atudenta of

peraonal time perapectives (e.g.. Bortner S Hnltach. 1972) have typically

assumed that the past-future differences which they have observed are

exciuaively due to changing persona! and aocia! realities. Eeintarpretation

of their results «ould clearly be called for if temporal aettlng alone were

found to affect judgment.

Considering the many waya in which people's perceptions of past and

future eventa have been found to differ (e.g.. gratfisch. Ekman. lundberg. *

Kruger, 1971; Cohen, 1967, Ekman S Lundberg, 1971, Koriat 4 Fischhoff,

1974), an effect of temporal setting on Judgment seems far from implausible.

Although this possibility has drawn little attention from psychologists.

It has been considered by philosophers and othera concerned with the study

of history. Probably the moat fre,uently raised contention in this prlmsrily

anecdotal literature ia that Judges sre more confident in desling with past

Ut. ■■ ■- " — mm • - ^.-..^-...-..^■^^-.■.■-^■^^^^i--^^--.^ —^ •■—'-■'"■'--"'-'-' : .■■-«-■--'^.*-»^'-"-''

- ■

IW^HiJi||p!IWHiWW«H

Tetrporal Setting

3

than with future events. Perhaps the most succinct expression of this

"sure past" hypothesis may be found in Seignobos and Langlois' classic

histuiiographic text, Introduction to the Study of History (1898):

The natural tendency is to forget in [the] reconstruction [of past

events] the results yielded by [scholarly] criticism, to forget the

incompleteness of our knowledge and the element of doubt in it. An

eager desire to increase to the greatest possible extent the art of

our information and the number of conclusions impels us to seek

emancipation from all negative restrictions. We thus run a great

risk of using fragmentary and suspicious sources of information for

the purpose of forming general Impressions just as if we were in

possession of a complete record, (p. 281)

Similar sentiments are readily found elsewhere (e.g., Ayer, 1956; De Jouvenal,

1967; Kissinger, Note 1; Pelrce, cited in Danto, 1965; Polak, 1965;

Toffler, 1970).

The source of reported differences in the judgment Df pagt an(}

future events is, however, often unclear. In each of the cases cited above

(see also Doob's, 1970, compendium), the change in the temporal setting of

the events in question is confounded with other changes likely to affect

judgment.

One frequent source of confounding is that the sets of past and future

events being judged contain markedly different members. Different events

should, reasonably, be judged differently. Even when similar events ^re

being considered, the judges Involved frequently possess outcome knowledge

about the past, but not about the future events; i.e., chey know how the

liiTilüWtltriirn' ^■ - ■ -—"^^-"-■^'"--:--' ■ ■■ - -... . . iilliiirii^iitrhill-MrtiMllitlBliiiWlMIIMiiilhiiltllt^^

^^^^^^^

Temporal Setting

A

former turned out. Such additional infor lation could reasonably affect

their perceptions. Although frequent, this latter type of confounding is

by no means necessary. There are past events whose outcome is unknown

and future events whose outcome is a foregone conclusion.

Even similar past and future events about which equal information is

available may be judged differently for reasons unrelated to their temporal

setting. Consider, for example, how the following events change in meaning

if they are set twenty years ago or twenty years hence: a trip to your

dentist, a parade of lesbians in the streets of New York, an earthquake on

Alaska's North Slope, writing ycur New Year's resolutions, or being unhappily

married in Italy (before or after enactment of the divorce law).

It is, of course, not changes in temporal setting which are responsible;

for these changes in meaning, but contemporaneous changes in relevant social

and personal realities. The world will be different in 1995 than it was

in 1955 and we, ourselves, will be different people in it. Temporal

relocation is no more responsible for changes in a mismatched Italian

couple's prognosis than geographical relocation per se would be responsible

for changes due to shifting the setting from Italy to New York.

The present study asks whether likelihood judgments for past events

differ from those for otherwise similar future events, with all these

sou-ces of confounding undone. Subjects in Past and Future groups were

presented with identical descriptions of events and asked to judge the

likelihood of possible subsequent developments, or "outcomes." For Past

subjects, the evaluated outcomes were set in the past; for Future subjects,

they were set in the future. These events were constructed so that temporal

setting-related changes in meaning were minimal.

■

^^^yy^Mfe^^Munl^BlgMgM^; «bgMM^ ^niiM^tiiriiBiiMritii m^^ ^mmmmmmmmmm

^^^^^^^^^^—^^^—^^-^-^^-^^^^^-^^—

Temporal Setting

5

Formally, deductive Inferences of this type are invariant with regard

to the temporal setting of the outcome judged. Thun, it should make no

difference to the judge whether the fate of a possible outcome (its occurrence

or non-occurvence) has already been decided (past events) or has yet to be

decided (future events)—unless, of course, temporal setting alone affects

his perception.

Two explicit hypotheses were tested: (a) past events are judged to be

more likely than equivalent future events; and (b) oast events are judged

more ex' -emely than equivalent future events. The former might be called

the "likely past" hypothesis, H.. , and the latter, the "extreme past" hypo-

thesis, H . H. refers to an increase in the central tendency of likelihood

judgments for past events; H refers to an increase in the dispersion of

likelihood judgments for past events. They e examined with outcomes

provided both by the experimenter (Experlmenu 1) or by the judge himself

(Experiment 2).

Experiment 1

Method

Design. In order to assess the effect of temporal sercing on judgment,

subjects were asked to evaluate the likelihood of a series of fifteen event

outcomes, each in the light of an accompanying event description. Theae

outcomes were set either in the past or future. The event descriptions were

always set in tue past. One outcome was evaluated for ea^h event description.

A specimen event description is:

■üiittn - - mmmmmt Jjjgtfg

■ ■

Temporal Setting

8

A French artist, Dollard St. Laurent, recently submitted a claim

to the government patent office. In this claim, he demanded a patent

for his artistic style. St. Laurent Is a well-known figure In

opposition circles.

The corresnondlne Past outcome question was:

In the light of these data, what. In your opinion, are the chances

that he received a patent?

The corresponding Future outcome question was:

In the light of these data, what. In your opinion, are the chances that

he will receive a patent?

In Experiment la, Past-Future differences were minimal, frequently

amounting to no more than a change in a single verb (as in the examples

given). No mention of temporal setting appeared in thr. instructions.

Experiment lb was identical to Experiment la except that the strength of the

manipulation was increased by a change of instructions and by adding details

which included tl;ue-related references. For example, the outcome presented

above was changed from ". . . he (the French artist) received a patent"

to ". . .he received a patent, and his success began a wave of similar

claims for receipt of patents from known and unknown artists alike."

Stimulus construction. The event descriptions dealt with a variety

of reasonably interesting content areas. The adjoining outcomes were

selected to evoke a full range of responses, from very likely to very un-

likely, over tasks. Several additional examples appear in Table 1. Nine

Insert Table 1 about here

of the 15 event descriptions which produced greatest inter-subject agreement within

groups and greatest inter-subject disagreement between groups in Experiment la

were reused.

i_^j||||i|||ttKi|a|H MMM^Mi -- - M^^^MiMM ***

Temporal Setting

7

The success of the experimental manipulation depends, of course, on the

independence of temporal setting changes from changes in accompanying social

realities. Among the precautions taken to guarantee this independence

were: (a) specifically setting events and outcomes in the near future or

recent past to eliminate possible differences caused by secular trends, such

as a rising crime rate or increasing sexual freedom; and (b) attempting to

equate the past and future time periods between the event and outcome, in

case probability of occurrence was related to the length of time allotted.

Each event appeared on a separate page. The different events were collated

into booklets with a standard order which varied the likelihood of sequential

outcomes.

Procedure. For each of the two experiments, the test booklets were

distributed to a single class of subjects from a pile in which booklets belonging

to the two conditions alternated. Given the minimal differences between the

versions and the minimum of disturbance during the administration, there is

little reason to believe that subjects were aware of the differences. This

procedure allowed distribution of the booklets to a fairly homogeneous group

of subjects in an unsystematic fashion ensuring fairly equal numbers of

subjects in each experimental condition.

instructions. In Experiment la. where every effort was made to obscure

the temporal setting variably the general instructions were identical for

all conditions. They read. "A number of short descriptions of social and

personal events appear below. Some are factual, others Imaginary. For each

description you will be asked to evaluate the likelihood of a possible outcome

nütMiiTiiir -■^-^-^»^«-^^--."--^■■'■..i^..^A:^i^^..s^.- ,....^..-^»-~.........^—^..^...-...■.^^,.,..Ja.^^^^i.^.^1,.^.^.-..^ MiiiiiiiiirtMMirillililiiliilt^l

^-^

ü*

Temporal Setting

8

In the light of the data. Indicate your answer with an 'X' In the appropriate

space." The twenty-one point response scale was labelled at three points:

20 = greatest possible chances; 10 = medium chances; 0 = least possible

chances. All Instructions and materials were in Hebrew.

For Experiment lb, they were changed to read: "A number of descriptions

of social and/or personal events which recently occurred (are about to terminate—

for Future subjects) appear below. For each description, you will be asked to

evaluate the chances that the event terminated (will terminate) in a particular

fashion."

Subjects. For Experiment la, the subjects were A8 students in an Intro-

ductory Statistics class at the University of the Negev, Beer-Sheba, Israel,

with 24 in each group. For Experiment lb, the subjects were 82 students

in an Intermediate Statistics class at the University of the Negev; 40 were

in the Past group and 42 in the Future group.

Results

H . Two measures used both here and in Experiment 2 are the "overall

media7 response" and the "median median response." The former Is the tLedian

of all responses by all subjects In a condition. The latter Is computed by

finding the median of each individual subjects' 9 or 15 likelihood judgments

and then finding the median of these medians. The median median is free of

the intra-subject dependence found in the overall median (to which each

subject contributed 9 or 15 responses).

The "likely past" hypothesis predicts higher likelihood assignments

to past outcomes. For both experiments, overall medians and median medians

were identically 10 for all conditions. Nor was there any tendency for Past

■ 3i ■

■ - -- ■"■'-:-"-J" ^■"■^-■^»-^-^-■^-^■I • - - ii' 11II i iii-itiiiitri-'^-"^--'-^^ji^^''^^^M°"

yiLiiyijJiiililijyiii. j.^jii|ipi|Hipii|pi^^

-

i:

Temporal Setting

9

outcomes of individual events to be judged more likely than the corresponding

Future outcomes.

H . The "extreme past" hypothesis predicts the assignment of more e

extreme (i.e., more disperse) likelihood values to Past outcomes. The mean

(absolute) deviation of each subject's responses from his median likelihood

assignment was calculated. The means of these mean deviations appear In

Table 2. Contrary to H , there is no appreciable difference in the dispersion

of Past and Future likelihood responses. Nor were there any consistent

Insert Table 2 about here

differences in the extremity (distance from 1.0) of the medi,--« likelihood

assignments to the Past and Future outcomes of individual events.

Discuaaion

Temporal perspective clearly fails to affect either the central tendency

or the dispersion of likelihood Judgments on the tasks used in Experiments la

and lb. If not completely illusory, the temporal setting effect must be

sought in more restricted and more carefully specified circur.stances. One

aspect of the present task which could have obscured such an effect is the

fact that subjects evaluated outcomes which they themselves had not produced.

Possibly, the act of production, absent in these experiments, elicits or

focuses attention upon critical differences between prediction and postdiction,

while the temporal setting of prepared outcomes is ineffectual. An analogy

with some appeal might be a difference between past (historical) and future

(science) fiction. Production of the two may evoke and require rather different

. ■„■,..;.,.:a.,^, ^.^...,.^J. ....^.....„«^.^.i^^.^^^^,*...^. ^.^^^^^^1^.^^^.^»^^M^J^..»^„.„:,„^..... t^^^ugimmmm

^«UIJIUII. (IWJJ. 1 l^^a^ltUlU ■U.<l,J,<HWMU«l«4IJl>VJW<IJ(k.!MI>Wl|P>JI|pn^^

Temporal Setting

10

"frames of mind." Once prodaced by others (authors), however, examples of

both genres may be Judged in much the same fashion by readers. Indeed, in

an unpublished study, Condry (Note 2; cited in Toffler, 1970) did find

differences in the length and imaginativeness of past and future outcomes

which pubjects had produced. Tversky and Kahneman (1973) have both hypothesized

and shown that likelihood judgmants may depend upon ease and conditions of

outcome production, with more ivailable outcomes judged more likely. Thus,

any temporal setting effect on production could lead to changes in likelihood

judgments.

Experiment 2 was designed to explore the possibility that outcome pro-

duction is necessary to evoke temporal setting-related differences in judg-

ment.

Experiment 2

Method

Design. For each of several event descriptions, subjects were asked to:

(a) produce a number of possible outcomes, and (b) evaluate the likelihood

of each outcome produced on a scale from 0 to 20. The Past group was asked

to.produce outcomes which may have happened, the Future group to produce

outcomes which possibly will happen. The event descriptions were identical

for both groups.

As little is known about how the conditions under which outcomes are

produced affect the way in which their likelihood is judged, several variations

on this basic design were included. In Experiments 2a and 2b subjects were

instructed to produce as many outcomes as possible; in Experiment 2c to

produce a fixed number of outcomes (four)—in the event that there were

effects which only emerge or vanish when subjects have exhausted their

^■^.^■....,...^..^..^MiiltlmiMgi,tgM ,^,.w—.,-,.».LJUl^.^..aamJ,.... .. -'■--"•■"-■'-ll1||i||Uf|i,|l^ ■„JJ..—t.^ ..,**

mmmmmmßm mum: uimmimmKm* mmmFirrmr*immaimmH***—--'fH

Temporal Setting

11

imaginations for possible outcomes. Experiments 2a and 2b differed in the

number of possible cutcomes which subjects could have produced, considering

the nature of the event in question. In Experiment 2a, this number was

extremely large (i.e., very many things could happen), whereas In Experiment

2b no more than ten or twelve outcomes were possible. It was thought that

the size of the set of possible outcomes might affect subjects' outcome

production and evaluation proeesses.

Each event appeared on a separate page. For Experiments 2a and 2b,

there were 12 numbered spaces for inserting and evaluating outcomes, along

with unnumbered space for an additional five or six outcomes. Following

the Instructions was a sample problem for which two possible outcomes were

affixed (In the proper tense) with likelihood estimates obtained from an

informal poll of ten students. The order of the events was randomized in

test booklets.

Stimulus construction. One of the event descriptions used in Experiment

2a was:

A well-known sculptor returned to his studio following a short vacation

and found that his former wife had burst into his studio during his

absence and destroyed statues which constituted all of his work for a

period of more than a year.

The second event was a Hebrew adaptation of that used by Condry (Note 2)

and concerned parents'handling a problem confronted by their adopted daughter;

the third appeared In Experiment 1 and dealt with a family crisis. Two

examples of the event descriptions used in Experiments 2b and 2c appear

in Table 1.

htiiiii'uri »V trnnirii iftlH'iÜrlMiaiiiM ■•'-' ^^^..^.U.^^.^,.,.^...-:^^.^..,.^-^.. .. ■■--'^"•'"■"""-"-'-ifii imiiiiri-i tfiliaiWWiliiiMiiiliiliiirili^iiriii IIM

HU .i.iiuMijn i \ i mmnm^Kw^mmmm^^'^mm wmmm nummMmmrnm^mf' ummiimMu. WWBMW^W

lemporal Setting

12

The events used in Experiment 2b also appeared in Experiment 2c. In

Experiments 2a and 2b subjects responded to three different event des-

criptions, in Experiment 2c to seven. Each spent approximately a half an

hour at the task.

Procedure. Same as Experiment 1.

Subjects. Sixty undergraduates in an Introductory Psychology class at

the Hebrew University of Jerusalem participated in Experiment 2a; 27 were

in the Future group, 33 in the Past group. Eighty-nine Introductory

Psychology students at the Hebrew University of Jerusalem participated in

Experiment 2b; 42 were in the Past condition, 47 in the Future. Forty-six

students in an Introductory Methodology class at the University of the Negev

participated in Experiment 2. They were equally divided be .ween the two

conditions.

Results

Power of manipulation. In Experiment 1 it was impossible to ascertain

whether subjects had actually attended to the temporal setting of the outcomes

they judged. The present design, however, affords a simple and effective

test: noting the tense of the verbs appearing in the outcomes produced.

Table 3 shows the proportion of subjects who used the wrong tense in each

condition. For each experiment, although most subjects did respond properly

Insert Table 3 about here

to the temporal setting instructions, there was a clear tendency to transform

poatdictive into predictive tasks (i.e., to use the future tense for possible

past outcomes). In each case, wrong tense responses were randomly distributed

"""" '""■■-^■"■■^•^ —,■»■....,.„-,>-.■., : ■ ■.■:...-.■..■ ..-.^^u,,.^.^^ i. - ^^Jgljlrcjgglggll^gl^ mlimmiJm^tjll^

■MiRPin^MMmv^ iv ^jmimmmmmmrii&m

Temporal Setting

13

over 8ub|ects, indicating that they were not produced by any subset of subjects

who "missed" the instructions but by a large number of subjects who tended

to "lapse" into prediction when .^ostdiction was required.

Rp Experiment 2a produced uome slim support for H . The overall

medians for Past and Future, respectively, were 12 and 10, 14 and 10, and

15 and 10, for the three events, ihe corresponding median medians for Past

and Future subjects were 12 and 10, 12 1/2 and 12, and 14 and 10. Thus,

in each case. Past outcomes were on the whole more likely than Future

outcomes. For the third event, this difference was significant (x2(l) = 7.38;

p < .01; median -est).

This support vanished in the remaining experiments. For both Experiments

2b and 2c, the overall medians and medians medians were essentially indistinguish-

able on the bcsis of temporal setting.

He. He predicts greater dispersion for past outcomes, indicating Past

subjects' greater confidence in their ability to distinguish degrees of

likelihood in the outcomes which they have produced. The most relevant

measure of the dispersion of subjects' likelihood judgnents is the mean

(absolute) deviation of individual subjects' likelihood judgments for the

various outcomes he had produced for each event. The means of individual

subjects' mean within-event dispersion appears in Table 4. As can be readily

seen, there are no consistent differences. Within-event dispersion was

Insert Table 4 about here

slightly larger for Past than for Future subjects in Experiment 2a, and

slightly smaller in Experiments 2b and 2c. None of these differences are

■ ---. Ül — mm m--**

,1II ILiM,lMMMlllJMÜliJ jj J mtmgmmmismmmmmmwmmmm ■!* 1 "■'»■»JJ»

1 empora] Setting

14

significant. Combining these results as suggested by Winer (1962, p. 44) k

yields z = -.65 ( -E t./Jk ; where k. •= number of groups), also Insignificant. 1-1

In no case does restricting attention to those events which had the

largest proportion of correct tense responses, restricting attention to correct

tense responses or treating future tense responses of Past subjects as having

been produced by Future subjects, and vice versa, provides any additional

support for either H, or H . 1 e

Effect of production. A number of analyses showed the Judged livelihood of

outcomes to depend upon the conditions under which they were produced. For

example, the later an outcome is produced, the less likely it is judged,

(the rank order correlations between likelihood and position of production

for individual subjects had a mean of -.26 over groups). The same relation-

ship was found within categories of similar outcomes produced by different

subjects (see below). Subjects who produced any given outcome earlier tended

to find it more likely (mean rank order correlation over outcome categories ■

-.17). In addition, the more outcomes subjects produced, the lower the median

of their likelihood judgments (mean rank order correlation between number

of outcomes produced and median likelihood judgment = w.27). The latter

relation suggests a possible artifactual source of the small Past-Future

difference which emerged in Experiment 2a. Future subjects might have assigned

lower probabilities as a result of having produced mbre outcomes. This is

not, howevrr, the case, as Future subjects actually produced slightly fewer

outcomes in both experiments where they were asked to produce as many outcomes

as possible (2a and 2b).

—■ —

___ ■JMlillli -»«

iiyawMMjMiuBPiiiPffiBWiiiiiww^ MWiiwiiwiiBpBiWiMywJMWWiw^^ J ^-'■Wt kMrnmrnnmymmmm

Temporal Setting

15

Outcome content. Outcomes were categorized according to content in

order to obtain further insight into Papt-Future differences. Two questions

asked were: (a) Do Past and Future subjects evaluate the likelihood of

similar outcomes similarly? an-l (b) Do Past and Future subjects tend to

produce different outcomes? The answer to the first question ..is a solid

"no," to the second a tentative "yes."

Regarding question (a), for the event in the example above, there was a

marked tendency for the medians of corresponding outcome categories to be

more likely and more extreme for Past than for Future groups. There were,

however, no systematic differences for the remaining two events of Experiment

2a, nor for the events of Experiments 2b and 2c. Regarding question (b),

the distributions of Past and Future responses over the outcome categories

were compared by chi-square tests following elimination of very small categories.

The two distributions were significantly different (p < .001) for all three

events In Experiment 2a, for one of the three in Experiment 2b, and for four

of the seven in Experiment 2c. Impressionistic analysis of these differences

indicated that t'ast outcomes tended to be more complex and imaginative than

Future outcomes.

Discussion

Contrary to the sure past hypothesis, there is no consistent evidence

that temporal setting affects either the central tendency (H.) or the

dispersion Oy of likelihood judgments, either for subject- or experlmenter-

... , . . . ^:.. ^A^.,. .,.■ ..»-.„.-ao^fc..»^*-.

ifliuiaramjwuwwww^ mmmmmm i^iiBjwfiimjiijppp mimmmmmmirimemmB^m

Temporal Setting

16

produced outcomes, for any of the varied events and tasks used here. The weak

effects obtained in Experiment 2a must, then, be considered a chance result in

light of the mass of contradictory evidence.

There is, however, evidence that temporal setting does affect the pro-

duction of possible outcomes—as reflected in: (a) the tendency to transform

postdictive tasks to predictive (i.e., the production of future tense Past

outcomes); and (b) the differing distributions of Past and Future outcomes

over categories for eight of the 13 events used. Temporal setting did not,

however, affect the number of outcomes which subjects were capable of

producing (Experiments 2a and ?b).

Aside from the evidence they provide on the effect of temporal setting

on likelihood judgments, these results provide some general insight

into how people perform such judgments. Tversky and Kahneman (1974) have

suggested two heuristics which judges might use in likelihood estimation

tasks like those used here; "availability" and "representativeness." The

availability-guided Judge deems an outcome likely according to the ease

with which it is produced or imagined. The representativeness-guided judge

evaluates the likelihood of an outcome by the degree to which it appears to

be a natural outgrowth of the situation described.

The fact that temporal setting affects outcome production without

affecting likelihood Judgments suggests that these likelihood judgements

rely on representativeness rather than availability—which should

be sensitive to changes in production. It is, in turn, highly plausible

that representativeness-guided judgment would be insensitive to temporal

setting. Not only is there no obvious way in which temporal setting would

" • ■■ --■■——-'-—- ^—--- | ^ö^^j^i^^^^ }matmmmmgMm

mmmmmmmmmm* \.mmmtmMtmrmmmmw..i- MUfiimmmmiBI mKtfmmHU

Temporal Setting

17

change the perceived evenc description-possible outcome fit, but it has

been found that reliance on npresentativeness obscures such selient and

normatively important aspects of event descriptions as sample size and

base-rate probabilities (Kahr.eman & Tversky, 1972, 1973),

What remains to ba explained is why some people believe in the ''sure

past" hypothesis. The simplest explanation is that pastness is confused

with its ecological correlates, concomitants which do, in fact, generally

lead to greater judgmental confidence. One correlate of temporal setting

which may contribute to a "sure past" illusion is contemporaneous changes

xn social realities. If human behavior is perceived to have been more

predictable and stable in the past than it will be in the future (another

claim advanced by Toffler), then pastness might well be misidentified as

allowing Judgmental sureness.

A second, possibly more significant, correlate of temporal setting

is outcome knowledge. Judges looking at the past typically know "how it

turned out." With this additional knowledge, and the hindsight it confers.

Judges are able to see and show the relative "inevitability" of past

events (Carr, 1961; Flschhoff, 1974; Fischhoff & Beyth, 1975; Nowell-Smith,

1970). Outcome knowledge seems to enable people to impose upon the past a

degree of certainty which they would not aare to imposa upon the future.

The relative prevalence of outcome knowledge for past events may create

the mistaken Impression that pastness alone is a sufficient condition

for uncertainty Judgment. Past and Future subjects here, however, were

equally Ignorant of the outcomes of the events they Judged.

A third variable related to temporal setting may be degree of control.

One can do something about the future, but very little about the past.

Virtually all of the literature concerning perception of past and future

....;.■ ■.■,..^:_T... --^ - ■ -- ■ j|||U|gfiiM|g||Utftt|Mb|u - - ■ - " ■-

'miwmmmmmgmKiKmKmmimmmmmmmmmimmmmm •mmmmmm wmmmmm wmmmm^mmmmmfif^

Temporal Setting

18

events (Dcob, 1970) deals with events In the Judges' personal rjaet or future.

Such personal Involvement may give the future a feeling of openness or con-

trollability which rnduces Judgmental confidence as it encourages phenomena

like "magical thinking" (Rothbart & Snyde., 1970). Again it would not be

pastn^ss, but a correlate which produces greater confidence.

In conclusion, it may be worthwhile to consider the general paradigm

which underlay the p.resent studies and which seems to offer possibilities

for fruitful interaction between psychology and other disciplines. A

common sense observation routinely put forth and used by a number of scientists

and Bhilosophers of science has provided the Impetus for investigation of a

systemir variable of general psychological interest. Empirical analysis

has, in turn, shown the inaccuracy of these anecdotal observations and

suggested needed refinements. It had also been found, anecdotal reports

to ':he contrary, that the temporal setting of events is a variable which

does not have to be considered in the design and comparison of judgmental

studies.

■ --— i i .mi ■■.■«■■■ ^M^MMUMI

mmmmmimmnmm m**nmi!mmmmmmmim*m**m*miwm»m*v!in"*»m ^jit-m... °*~'*™mii.mmmmwgwmmm*immi*mr*t*'mmmmmmmiim*i'***'**

1.

2.

Temporal Setting

19

Reference Notes

Kissinger, H. Opening speech at the Geneva convention, December, 1973.

Condry, J, Personal communication, October 3, 1973.

- —-— mmm*mumm*M

f j ,i.A^m.^j.a?yjBH^yi.|i|iija^iyiL^^^

Temporal Setting

20

References

Ayer, Ä. J. The problem of knowledge. London: MacMillan, 1956.

Bortner, R. W., & Hultsch, D. F. Personal time perspective in adulthood.

Developmental Psychology, 1972, 7, 98-10:.

Bratfisch, 0., Ekman, G., Lundberg. U., & Kruger, K. Subjective temporal

distance and emotional involvement. Scandinavia Journal of Psychology,

1971, 12, 147-160.

Carr, C. E. What is history? Hammondsworth: Penguin, 1961.

Cohen, J. Psychological time in health and disease. Springfield, 111.:

Thomas, 1967.

Danto, A. Analytic philosophy of history. Linden: Oxford, 1965.

DeJouvenal, B. The art of conlecture. New York: Basic Books, 1967.

Doob, L. The patterning of time. New Haven: Yale, 1970.

Ekman, G., & Lundberg, U. Emotional reaction to past and future events

as a function of temporal distance. Acta Psychologica, 1971, 35, 430-441.

Fischhoff, B. Hindsight and foresight: The effects of outcome knowledge

on judgmer-- under uncertainty. Oregon Research Institute Research Bulletin.

1974, 14, No. 14.

Tischhoff, B., & Beyth, R. I knew it would happen—remembered probabilities

of once-future things. Organizational Behavior and Human Performance,

1975, in press.

Kahnet.an, D., & Tversky, A. Subjective probability: A judgment of repre-

sentativeness. Cognitive Psychology, 1972, 2» 430-454.

Kahneman, D. , & Tversky, A. On the psychology of prediction. Psychological

Review. 1973, 80,237-251.

m —— ---— mmmutmm

PlW^^jllW^^JiJ^IM-ll^lPMIIllUW. jpj^WjJW.JilllP^PPpW'! mmmmmismmmjm'm ■ppppipww

l.^.-,.-. :..L ._,,,

Temporal Setting

21

Korlat A., & Fischhoff, B. "What day is today?" An Investigation of the

process of temporal orientation. Memory and Cognition. 1974.. 2, 201-205.

Nowell-Smith, P. Historical explanation. In H. Kiefer & K. Munltz (Eds.).

Mind, science, and history. Albany: State University of New York Press,

1970.

Poiak» F- Prognostics. Amsterdam: Elsevier, 1965.

Rothbart, M., i, Snyder, M. Confidence in the prediction and ppstdiction

of an uncertain event. Canadian Journal of Behavioral Science. 1970,

2, 38-43.

Seignobos, C, & Langlois, C. Introduction to the study of history. New

York: Holt, 1898.

Toffler, A. Future shook. New York: Random House, 1970.

Tversky, A., & Kahneman, D. Availability: A heuristic for Judging frequency

and probability. Cognitive Psychology. 1973, 5, 207-232.

Tversky, A., & Kahneman, D. Judgment under uncertainty: Heuristics and

biases. Science. 1974, 185, 1124-1131.

Winer* B- J- Statistical principles in exEerimgntaj design New York:

McGraw-Hill, 1962.

iimrrm |iM-i_Ml4 *ämaiiiiäimimmiä^lä*mäkmtiäimMx MMWUiUWWMiliiMÜ - '—"^ irr'ii'iiniilü

npünp mvimimmmmmmmmmm muMmmimiimmm, mmM^mm.-Mwmmmwm.,'-'j'.*&^*mmmfi*i'^

Temporal Setting

22

Footnotes

This study was supported by the Advanced Research Projects Agency

of the Department of Defense (ARPA Order No. 2449) and was monitored by

ONR under Contract No. N00014-73-C-0438 (NR 197-026). The research reported

constituted a portion of a doctoral dissertation submitted to the Hebrew

University of Jerusalem, May, 1974. Consultation with Amos Tversky, Paul

Slovic, and Sarah Lichtenstein contributed substantially to this research

and is gratefully acknowledged. Reprint requests may be sent to the author

at Oregon Research Institute, P. 0. Box 3196, Eugene, Oregon 97403.

A third group which received future outcomes (as above) with future

event descriptions (e.g., "A French artist is about to submit a claim . . .")

was also included in this study. It allowed evaluation of the effects of

changing the temporal setting of the event description as well as the

outcome. Results for this group are not reported in detail here as it was

eliminated from further experiments because of subjects' reported discomfort

with future tense f^vent descriptions. Responses to these items were in-

distinguishable from responses to the corresponding Past and Future Items.

2 Additional evidence of the dependence of likelihood judgments on

aspects of outcome production emerged in Experiments 2b and 2c below. It

is omitted for the sake of brevity here, but is available upon request.

«■ i ■ - " ■—"^ «naiir- - iMiiiiii^aimn 11 ■ - ■ ■ ■ - ■ ■ —-^ ■■■■

WIBJIJ.HUJIU! illil m*ma ■ i —mmm Hiuiawiiwu i tmm^mmm*!'mimi*r^'****'mmmi*mi^*m

5 4-1 u 0) in

U o I

TS *-> c n) Ai cd S

1 0 J3 n) -H •H £ w (U > j: M & O. tu Q) T3 to ^ o

I •> 0) x> u • •H -a -o m js H -H H 35 A! 4J

q <u o a to M (0 n) oi o c d) -H

M T3 3 -H 0.,fl 0) to (0 ja H «^-^ *J H tn H M C T3 O

c <u 0) n i-i ^ •l (TJ tn ^ (*-i 3 IM

M u u a o ■u 0» m e8 <H 5 ja c j: H tu co 0) M (U -H M 13 C (U -H Jl H to c i-t x: tu to öO p. n) 4J j: ß

• (U ^ U 4J w ü ci > cd s n) -H tu o xi c H U H J2 4J 4J -H

1 1 T3 0) 0) 01 J3 ,o ,Q M O

09 ta tu 01 H -H rH -H *J tu -H J3 H J3 C a 60 •H •H 3 (0 c S to 3 ta o •H '-' cd ^ cd o <u u

G U 0) e J= C J3 0) IU 4J cd co Cd !0 > a ÜOiH 00-H U tu 0) tU iH 0) H 3 co ja 60 ja 60,0 c

c a ^-s O T3 W • •a w • l-i (X H tu iH 0) tU 3 •H J4 60 •H .Ü 60 W J3 cd 2 ja cd cd a to c

3 tU <U Ü 0) 3 CJ 0) 3 a 60 a öo O -H 4J

0) to Q lU co C u u u JS cd jz cd C H Cd 4J o rH 4-> O H (U 3 60

4J 4J O U • *J • 4J iH -H 0) /■^ to ^-\ CO ^1 »4-1 -O • c U • (3 U •H »w c •H •H •H -H > -H «H

• Ö0i*-i • 60U-( w-a j,.

ta ta tu

i to a o 4J ^ td -H c •a cd O J2 td a Ä J3

i4-i T3 to J: n) 4J CJ tu ß u Cü

Cd N 0 •i <U 0) •H -H tn rH H l-i

•> H P--H § ja M td B x: •H ß

X) tu tu cd •rl tn tu tu 4-i M x: 4-i ß tn > tn ß u cd ■g o tu ta ■H •> J3 a

M ß H 4J 4-1 ■H p. o td O) CO o • rH rH CO -H ß T3 M cd ß to 3 a o tu tu & H

H i 4-1 60 B tn x: » "O cd •H c ß td 4J o 4J rH ß

0) 4-J 0) •H x: cd cd ß •H 3 tH -H 02 E 60 O ß 0 •H O U-i ja •H cd u s 4J ? td IM M U <ü tx cd tu H O tu ß a) x: a 0 Xi 0) J3

p- ■< S 4J cd 4J 4J x: 4.) to X 0) w

i 1 >^>4H -« ß ^ ß ß IU Ü 0 u o 0) rH

•H -H •H • H •o cn PL. to p. C0 e e rH Cd

/"■>> cd /-N cd H ja tu x: tu x: •H ja o ja o S tu

^ ja • rH (U rH IU 13 H Xi H x: •O 4J o •H 4J •H u tu cd o 3 & •H ja 60

v^ IM ß 4J o o 0) ^

in co 13 >•. O cd tu cd tu rH MH » ti U B 4J rH

iH •H 3 cd tn CU 4-1 0) 4J ja u u

J2 ja •H iH IU • 0) •> 4J 4-1

• x: to • x; CO Cd 0) 4J tt •u P.ja rH

■ •H • •H axi U Xi 'J x: B 4J

• cd to • cd co tu cd

0) i ja

C0 4-1 0)

I tn

r-l to td X3 tu 4-1 u

T) td <U

X) 4-4 •H o O

IU u X) IU 60 B cd C ß ■H Sj IM c

01 tu 4-1

e cd 4-1

to 4-1 4-1 •H. O

ß 13 ß ß td tH p. 60 X IU tu ja

ß

s

n)

ja tn to ß o >-,-H ß 4-1 td td fr u e tu O P. o o

>. • 4J 10 u tu o) B p, o o x: u P-13

rH ß -H

•H 3 ja

H cd O 0) 4J

T> O

o to 4J rH

cd

i-l 4J ß 3 o ja

4J *J ß •rl cd cd ß ja d.

•H 4-) B o

o) p, u ja o

rH 0) 4J <4H x: CJ 4J tu cd •n T3 o x: oj MUß p. 3 m

cn o 4-1 CM co co to k cd ß •H > cd

13 OJ tu ja 4-i H cd

o u u cd P. 0)

B ^ O 4J i-i cd

M-i xa 4J

4J C Vi tu o B >4H 0) 4-1 tß cd cn 4J O CO rH

cd

60 13 3

0) B o u

3 O

tn u tu

o tu ja

cn 4J

^/-s Wi ß 01 tu cd u 61 P.-H tfl B «w ß o CJ rH i

rH 0) -H (U ja & >

• 13 U 0) 01

• t-i IM •H >4H

. <M (U

c

•H u p. X w

g^ g^ O IU

ja t0 4J

ß 0) td >-i P.-H B MH o O rH

H IU iH ja »

0) • M

•H • IM

4J ß IU B

•ri U 0)

w

13 tU ß

rH O rH rH -H rH OJ tn -H

%%£ IU 01

^ in ß 0) -H -H 6oja o Sp.

/-N P. cd rH cd B 0)

O "O 0) ß ß > cd cd

•H U 4-i * 0 t- m 01 H ß

UH -H Cd MH S rH CU v^ p.

IU 0) ß co o o

p- U P, 0) o 60 cd 13 ß T)

ß 0)

cn tu cd 13

•H • ^ cd c u ra ß tn rH •H -H &. o ja O 0) u.>f4 xa CO O 4J

—'-— ^J^—-^■..■J.^. t „--■ ......»...„■v^.^....-^,-^. iwiriliftiif'rlliiiniiiTiitntiiiiiim'irrtiiiiriTiirr-n ^'--•'■'■'■-■-^-■■■^^^^^*^^..^...„.^ ^"'■'^"-^-'t'■-'"-'■ -

ftillJ4W4WJBIIIHimilipB|li«WMJUtLJ wmm mtß'ntmmfwm mm ^WJ*A'&!wmvw*M1!M% l*.J ^M^J*1^

I '

:

5 (U

O

<u H

CM

o CM

s ja

i •H U 0)

>. M O H 0) w ^S (0 H U

o wo u

0) 01 3 -• Ü0 3 0 a) &0 o* C O M tO 0) •a U a) MTI T) H

'o >>H 4 M 0) t) M t-H rH 01 •

•> (d 4-1 i-t 1-1 Xi 0) n S 4J J3 a) 4-i • ■o M 0) U a) a •H O M-l rn ^ m o CJ 4-1 O 0) (d -H 0) U <U 43 o _ *J T) n) w M 4-) & e o

M (U 4J o) n) o JS « & IW J3 4J 4-10) O cd 4-1 ^-1

2 ^SK. 0

T3 -Ö 4J

a <u a) CDS 3 « QJ r-l *J cd 4J O Ü • > ,Q W Ö T3 o

ij a) o g m M 0

>> cd o) o ^ <u 0)

cd o- ^ M A U U o> 4-1 <4-l 4-1 <d a) rH ^ O 0 «5 4J «d -U rH (U (0 o

U-i tD O M -H 4-1 x) rt <U M cd JC Ö tH O U fi- a td a) (4-4 cd u QJ

4J a) a) u o > Ä rt Ä w >. a •* •H a. M 4-i P- JS u M a) a) a) Q> U U 0) Put-< o > -H 0) en O 0) H X Hj a) <J <u

•H a) a) n) u-i Z »-I s

M M O Q) O "O O M-l •H "O «W >. <U >4-< tS TJ o o c

u cd

o 3 1-1 o f*.

a •H Xi

o •a •o c

cd co 4J cd ,0 iH 0 J3 o ja a» M O O a c a) 4J

•H U CO u » •H T3 0) 3 CO Ä a» M O 4-1

o -a ,o O •> cd 3 cd <U

T3 rH 1-1 1-1 c a. a a) o) -H ■i ^1^^ CO

MO) C 0) o) tn O Ä

J3 O • «H • 4J 4J •-) 4J 0) fs O C -H c ^

<u a c o) > g o)

o ,a •H

J: cd 4-i "ö CO •n ti j: cd •H (U

a) co U -a Ö O H iH (U •H Ö 4J 4-1 ^3 Tl >

o u a

01 u (1) »

44 d

0) Ä 4J

0) M cd

* C3 o

•H C

•H P. O •

to u 3

§•3 >. a

■H c 4J

1-1 W

• J3 4J o cd cd « 0) U

14-1 ft o cd 4J c cd o •ö •H

(0 0) U Ul 0) 0) > ja 4-1 0)

M <4-l 3 o 4J

3 U fn x: bO 0)

•H Ä rH 4J

01 C x: •H 4-1

Tl c 0) M M g cd

0) P.

• ■ P B cd cd bt to <u OJ

J3 n 0)

4J 4= c *J 0) C a 01 0) u 4J cd cd p 4J tn c

•H 0) B to o 0) tl to 4J CO 3 M c J2

P- .d u 01 cd JC

ü^u^üasäikä^iUiiiiäslüUüiyi J^--..»^^—.,..^„^.,.-.,,,., .ni->|..lT..^ „.-J...:,,. ..^..--.,J..A.,.^ ..J«^.-.-^^^-^...,... —^- —.. ■ - '-' —

•^mmammmmmmmmmmvmammmmmmimimimHftiimm

Table 2

Mean Between-Event Dispersion

Experiment

la lb

Condition Past

Future

3.18

3.57

4.76

4.57

t

df

-1.45a

46

0.68

80

Temporal Setting

25

Negative sign for t test for difference indicates

results contrary to H .

^^^^^...■,:...-.J,..^^.aa^tMi^ajit^^i^M»^««>«^ji^^»A^^.-J ^.■■^^„^^-^^^^.-„-^^^^»^»^■^^^^aa^,.»,^^—■,^..^M.M Hi.-ni.fiima "-■- -■ '--^-~'-J^-..^--^ . -■-,,,..*jm

IP"""« mmmmt ff*mjlggmmgi/emmi^*i^*miJLL,iim-+ 41JJI.UIJIIJHUILULI,...I a^mmmmmmimrw

Tense Produced

Temporal Setting

26

Table 3

Tenses of Outcomes Produced

Tense Required (condition)

Experiment 2a Experiment 2b Experiment 2c

Past Future Past Future Past Future

Past 92.9% 0% 78% 4% 89.9% 9.2%

Future 7.1% 100% 22'* 96% 10.1% 90.3%

a z 1.96 3.97 0.66

aThe difference In the proportion of subjects using the wrong tense

In each condition.

\\\tutt'\ ihi*ii i

lu tmmmmmmmm. 'T^i

o

§ •H 01 u 0) p. 0)

0) H

H

•H

p.

(0 CM

vD

fl

CM

ro

r~. CM Oi og lO sr • • • CO cn o

00 o\ •* vO ON vO

vO

o CO

m oo

e 0)

CM

u-i

ro

ON -3-

CM

PL,

vO

CM ON

CM

(U U 3 4J 3

o

o I

o

o o

I

o I

u-l in

eg i - oo

in o oo r- m oo r* oo • • • CO CO o

I

cd 00 CM r*. r~- >* vO «» oo

• • • < •* o

00 iH PO oo h« r^ o m

en m

oo m

oo

m o

u n) u u a o u

0)

u m <u u cd o •H

a

0) u c 0)

(1)

•a

u o

a)

u o

4-1

&

0) >

■u cd 60

---- MMM ---■ -^ ^.^^ ^^^^^ —■"-^-^-- I — -.„.^ ..