ArticlePDF Available

Putting Adjustment Back in the Anchoring and Adjustment Heuristic: Differential Processing of Self-Generated and Experimenter-Provided Anchors



People's estimates of uncertain quantities are commonly influenced by irrelevant values. These anchoring effects were originally explained as insufficient adjustment away from an initial anchor value. The existing literature provides little support for the postulated process of adjustment, however, and a consensus that none takes place seems to be emerging. We argue that this conclusion is premature, and we present evidence that insufficient adjustment produces anchoring effects when the anchors are self-generated. In Study 1, participants' verbal reports made reference to adjustment only, from self-generated anchors. In Studies 2 and 3, participants induced to accept values by nodding their heads gave answers that were closer to an anchor (i.e., they adjusted less) than participants induced to deny values by shaking their heads--again, only when the anchor was self-generated. These results suggest it is time to reintroduce anchoring and adjustment as an explanation for some judgments under uncertainty.
Research Article
VOL. 12, NO. 5, SEPTEMBER 2001 Copyright © 2001 American Psychological Society 391
Differential Processing of Self-Generated and Experimenter-Provided Anchors
Nicholas Epley1 and Thomas Gilovich2
1Harvard University and 2Cornell University
Abstract—People’s estimates of uncertain quantities are commonly
influenced by irrelevant values. These anchoring effects were origi-
nally explained as insufficient adjustment away from an initial anchor
value. The existing literature provides little support for the postulated
process of adjustment, however, and a consensus that none takes place
seems to be emerging. We argue that this conclusion is premature, and
we present evidence that insufficient adjustment produces anchoring
effects when the anchors are self-generated. In Study 1, participants’
verbal reports made reference to adjustment only from self-generated
anchors. In Studies 2 and 3, participants induced to accept values by
nodding their heads gave answers that were closer to an anchor (i.e.,
they adjusted less) than participants induced to deny values by shak-
ing their heads—again, only when the anchor was self-generated. These
results suggest it is time to reintroduce anchoring and adjustment as
an explanation for some judgments under uncertainty.
In what year was George Washington elected president? What is
the freezing point of vodka? Few people know the answers to these
questions, but most can arrive at a reasonable estimate by tinkering
with a value they know is wrong. Most know that the United States de-
clared its independence in 1776, so Washington must have been
elected sometime after that. And most know that alcohol freezes at a
lower temperature than water, so vodka must freeze at something
colder than 32 F. To answer questions like these, in other words, peo-
ple may spontaneously anchor on information that readily comes to
mind and adjust their response in a direction that seems appropriate,
using what Tversky and Kahneman (1974) called the anchoring and
adjustment heuristic. Although this heuristic is often helpful, adjust-
ments tend to be insufficient, leaving people’s final estimates biased
toward the initial anchor value.
To examine this heuristic, Tversky and Kahneman (1974) devel-
oped a paradigm in which participants are given an irrelevant number
and asked if the answer to a question is greater or less than that value.
After this comparative assessment, participants provide an absolute
answer. Countless experiments have shown that people’s absolute an-
swers are influenced by the initial comparison with the irrelevant an-
chor. People estimate that Gandhi lived to be roughly 67 years old, for
example, if they must first decide whether he died before or after the
age of 140, but they estimate that he lived to be only 50 years old if
they must first decide whether he died before or after the age of 9
(Strack & Mussweiler, 1997).
Anchoring effects have been demonstrated in numerous contexts,
including the evaluation of gambles (Carlson, 1990; Chapman &
Johnson, 1994; Schkade & Johnson, 1989), estimates of risk and uncer-
tainty (Plous, 1989; Wright & Anderson, 1989), perceptions of self-
efficacy (Cervone & Peake, 1986), anticipations of future performance
(Switzer & Sniezek, 1991), and answers to general knowledge ques-
tions (Jacowitz & Kahneman, 1995). Anchoring and adjustment has
also figured prominently as an explanatory mechanism underlying such
diverse phenomena as preference reversals (Lichtenstein & Slovic,
1971; Schkade & Johnson, 1989), probability estimates (Fischhoff &
Beyth, 1975; Hawkins & Hastie, 1991), trait inference (Gilbert, 1989;
Kruger, 1999), language production and comprehension (Keysar &
Barr, in press), and various egocentric biases such as the spotlight ef-
fect (Gilovich, Medvec, & Savitsky, 2000) and the illusion of transpar-
ency (Gilovich, Savitsky, & Medvec, 1998).
Anchoring effects have traditionally been interpreted as a result of
insufficient adjustment from an irrelevant value (Tversky & Kahne-
man, 1974), but recent evidence casts doubt on this account. Instead,
anchoring effects observed in the standard paradigm appear to be pro-
duced by the increased accessibility of anchor-consistent information
(Mussweiler & Strack, 1999, 2000; Strack & Mussweiler, 1997). The
attempt to answer the comparative question—say, whether Gandhi
lived to be 140—leads an individual to test the hypothesis that the ir-
relevant anchor value is correct—did Gandhi live to 140? Because
people evaluate hypotheses by attempting to confirm them (Crocker,
1982; Snyder & Swann, 1978; Trope & Bassok, 1982), such a search
will generate evidence disproportionately consistent with the anchor.
The absolute judgment is then biased by the evidence recruited in this
confirmatory search. This alternative account, accompanied by fail-
ures to demonstrate a process of adjustment, has led some researchers
to conclude that “anchoring occurs because of biased retrieval of tar-
get features,” and not because of insufficient adjustment (Chapman &
Johnson, in press; see Mussweiler & Strack, 1999, in press, for a dis-
cussion of anchoring effects with and without adjustment).
We believe this conclusion is premature. In particular, we suggest
that just as memory research was sidetracked by an overly persistent
analysis of people’s ability to recall nonsense syllables, so too has an-
choring research been sidetracked by an overly persistent analysis of
people’s responses in the standard anchoring paradigm. Outside this
paradigm, anchors are often self-generated, rather than provided by an
experimenter or other external source. People know George Washing-
ton was elected after 1776, but how long after? People know that
vodka freezes at less than 32 F, but how much less? Externally pro-
vided anchors, even outrageous ones, differ from self-generated an-
chors because they have to be taken seriously, if only for a moment.
Self-generated anchors, in contrast, are known—from the beginning—
to be wrong. There is thus no cause to consider whether the anchor
value is correct and thus no engine of heightened accessibility of an-
chor-consistent information. This difference led us to propose that an-
choring effects are produced by insufficient adjustment rather than
selective accessibility when the anchor is self-generated. We investi-
gated this possibility in three experiments.
Address correspondence to Nicholas Epley, Department of Psychology,
William James Hall, Harvard University, Cambridge, MA 02138; e-mail:; or e-mail Thomas Gilovich:
Anchoring and Adjustment
392 VOL. 12, NO. 5, SEPTEMBER 2001
In our initial exploration, participants verbalized their thoughts
when answering questions involving self-generated and experimenter-
provided anchors. We predicted that participants would describe a pro-
cess of anchoring and adjustment only when anchors were self-gener-
ated. In these cases, we expected that the verbal reports would typically
begin with a reference to the anchor value, followed by a statement de-
scribing adjustment away from it (e.g., “The United States declared its
independence in 1776 and it probably took a few years to elect a presi-
dent, so Washington was elected in . . . 1779”). In contrast, we ex-
pected experimenter-provided anchors to produce little or no mention
of either the anchor or adjustment, consistent with the selective-acces-
sibility account of anchoring effects in the standard paradigm (Strack
& Mussweiler, 1997).
Fifty Cornell undergraduates were each asked four questions. Two
questions were ones for which most participants could be counted on
to generate a particular anchor value (e.g., “When did the second Eu-
ropean explorer, after Columbus, land in the West Indies?”—1492),
and two involved anchors provided by the experimenter (one high
value and one low value; see Table 1).
Participants were asked to explain how they arrived at the answer
to each question. Their responses were recorded, transcribed, and
evaluated by two raters who were unaware of our hypotheses. For
each response, the rater evaluated whether the participant appeared to
know the anchor value, used the anchor as a basis of the answer, and
mentioned adjustment from the anchor to arrive at a final estimate. In-
terrater agreement was .94. A third rater who was also unaware of our
hypotheses resolved disagreements. Participants were considered to
have utilized anchoring and adjustment only if their verbal reports re-
ferred to both the anchor and a process of adjustment.
Results and Discussion
As predicted, participants were more likely to describe a process of
anchoring and adjustment when the anchor values were self-generated
than when they were provided by the experimenter. Of those partici-
pants who appeared to know both self-generated anchors (n
94% made reference to anchoring and adjustment in response to at
least one of the self-generated items, and 65% did so in response to
both. In contrast, only 22% of the participants (n 50) described an-
choring and adjustment in response to at least one of the experimenter-
provided anchors, and only 4% did so in response to both (see Table 1).
To assess the statistical significance of these results, we calculated the
percentage of items for which participants reported a process of anchoring
and adjustment for the self-generated and experimenter-provided items.
Four participants were excluded from this analysis because they knew
neither of the self-generated anchors. As predicted, participants were far
more likely to report using anchoring and adjustment when considering
self-generated anchors (M 73.9%) than when considering experi-
menter-provided anchors (M 13.0%), paired t(45) 8.56, p .0001.
These results indicate that self-generated anchors activate different
mental processes than experimenter-provided anchors. One might be
concerned, however, about relying on participants’ self-reports given
the widespread doubts about whether people can accurately report on
their own mental processes (Nisbett & Wilson, 1977). One might also
be concerned about a Gricean alternative interpretation of these find-
ings. That is, participants may have been less likely to mention the ex-
perimenter-provided anchor value and how they adjusted from it
because the anchor value was already mentioned in the initial compar-
ative question. Note that this interpretation is rendered less plausible
by the fact that the same pattern of results was obtained when we
scored participants’ responses for statements of adjustment only,
rather than statements of the initial anchor value and adjustment. Nev-
ertheless, we conducted the following studies—which manipulated
the process of adjustment rather than assessing it—to rule out this ex-
planation completely.
When people adjust from self-generated anchors, they may do so
in one of two ways. One possibility is that people “slide” along some
mental scale, continuously testing until they arrive at a satisfactory fi-
nal estimate. More plausible, we believe, is that they “jump” some
amount from the anchor—analogous to a saccade in reading—to a
more reasonable value and assess its plausibility. If the new value
Table 1. Percentage of participants describing a process of anchoring and adjustment in
Study 1
Question n
Percentage describing
anchoring and adjustment
Self-generated anchor
When was Washington elected president? 42 64
When did the second European explorer,
after Columbus, land in the West Indies? 37 89
Experimenter-provided anchor
What is the mean length of a whale? 50 12
What is the mean winter temperature in
Antarctica? 50 14
Note. Although 50 participants were included in this experiment, the number of participants varies for the
items with self-generated anchors because not all participants knew the relevant anchor (i.e., the date of
Columbus’s arrival in the West Indies or the date of the Declaration of Independence). The anchor value
provided was 69 ft for the mean length of a whale and 1 °F for the mean winter temperature in Antarctica.
Nicholas Epley and Thomas Gilovich
VOL. 12, NO. 5, SEPTEMBER 2001 393
seems plausible, adjustment stops. If it does not seem plausible, a new
jump or saccade is made, the new value is assessed, and so on.
Regardless of the continuous or discrete nature of adjustment, any-
thing that influences participants’ thresholds for accepting or denying
values that come to mind should influence the amount of adjustment.
If a participant is more willing to accept values, he or she will termi-
nate the adjustment process more quickly and provide a final estimate
that is closer to the original anchor value. If a participant is less ac-
cepting, he or she should continue to adjust and arrive at a final esti-
mate further from the anchor.
We sought to influence participants’ thresholds for acceptance or
denial by using the tried-and-true influence of motor movements on
attitudes and persuasion (Cacioppo, Priester, & Berntson, 1993; For-
ster & Strack, 1996, 1997; Martin, Harlow, & Strack, 1992; Priester,
Cacioppo, & Petty, 1996). Previous research has demonstrated that
people are more likely to accept propositions when they are nodding
their heads up and down than when they are shaking them from side to
side (Wells & Petty, 1980). We reasoned that asking participants to
nod their heads would make them more willing to accept values that
initially came to mind, and thus produce less adjustment from self-
generated anchors. Shaking their heads from side to side, in contrast,
would make participants more willing to reject values, and thus pro-
duce more adjustment from self-generated anchors. Because of this
difference in adjustment, we also predicted that participants would
generate an answer more quickly when nodding than when shaking
their heads.
Because nodding and shaking should not systematically influence
the selective accessibility of anchor-consistent information, we pre-
dicted these head movements would not influence answers to ques-
tions with externally provided anchors.
Participants (n
50) were told that the experiment was a study of
product evaluations, and that they would be asked to evaluate a set of
headphones while moving their heads from side to side or up and
down in order to assess the headphones under everyday use.
All participants listened to a tape containing 16 anchoring ques-
tions. To justify this procedure and reduce suspicion, the experimenter
explained that she wished to examine “implicit evaluations” that people
“form without conscious intention or effort.” She thus needed to busy
participants with another task while they were evaluating the head-
phones, in this case by answering the questions on the tape. Depending
on a random schedule, participants were then asked to nod their head
up and down, shake their head from side to side, or hold their head still.
The experimenter, who was unaware of our hypotheses, provided a
brief demonstration of the desired head movement for each partici-
pant, situated herself behind the participant, readied a stopwatch, and
began the tape. She recorded the answer to each question as well as
the time required to generate each answer.
The 16 anchoring questions, which were presented in a fixed order,
were divided into blocks of 4. In order to maintain the cover story, the
experimenter stopped the tape after each block and asked the partici-
pant to evaluate the headphones. All questions in the first three blocks
involved self-generated anchors (e.g., “How long does it take Mars to
orbit the sun?”—365 days), with participants completing one block
while randomly performing each of the head movements.
The last block contained four anchoring questions taken from Ja-
cowitz and Kahneman (1995). Participants repeated the head move-
ment made during the first block, and the experimenter recorded their
answers and reaction times to the comparative and absolute compo-
nents of each question. Because we were interested in adjustment, and
not anchoring effects per se, we did not manipulate the experimental
anchor for each question. We selected the four items that produced the
largest anchoring effects in the study by Jacowitz and Kahneman, and
provided high anchor values for two questions and low anchor values
for the other two (e.g., “Is the population of Chicago more or less than
200,000? What is Chicago’s population?”).1
Following this procedure, participants completed a questionnaire
that asked directly about the intended anchor value for each item in-
volving a self-generated anchor (e.g., “In what year did the United
States declare its independence?”) and whether they had considered
this value when generating their answer.
Results and Discussion
Two preconditions had to be met for an adequate test of our hy-
potheses about self-generated anchors. First, participants had to know
the self-generated anchor. Second, they had to report considering the
anchor when making their estimate. Participants who did not meet
these preconditions were excluded on an item-by-item basis. On three
questions, fewer than 30% of participants met both preconditions,
generally because they did not know the intended anchor value. In
some cases, this left no participants in one or more of the experimental
conditions. We therefore dropped from the analyses three questions
(about the fastest mile, death of the first apostle, and orbit of Io). This
left nine questions with self-generated anchors (three in the first block,
four in the second, and two in the third).2,3
To determine whether head movements influenced participants’ re-
sponses, we converted answers to each question to standard scores and
then averaged across all items within each head-movement condition.
Reaction times were logarithmically transformed to reduce skew, then
standardized and averaged in the same fashion. Answers to the four
items that required downward adjustment were reverse-scored so that,
as for the other questions, higher scores on this index reflected a larger
discrepancy between the anchor and final answer. As can be seen in
Table 2, a repeated measures analysis of variance on this composite
measure indicated that participants’ head movements significantly
influenced their answers to the items with self-generated anchors,
1. A complete list of the questions used in all experiments can be obtained
from the authors.
2. On two of the remaining items, the gestation period of an African ele-
phant and the orbit of Mars, not all participants adjusted in the right direction
from the intended anchor (i.e., some believed Mars travels around the sun in
fewer than 365 days, or that an elephant’s gestation period is less than 9
months). Because we were interested in the process of adjustment, we used the
absolute difference between the reported anchor and final answer on these
items. Higher numbers on these adjustment scores indicate a larger discrep-
ancy between the anchor and final answer, or larger adjustment. The same pro-
cedure was employed on all of the items with experimenter-provided anchors,
for the same reason. Note, however, that the results in both this study and
Study 3 were unchanged when participants’ raw estimates, rather than adjust-
ment scores, were used for the items with experimenter-provided anchors.
3. Some participants confused Fahrenheit with Celsius, reporting 100
the boiling point of water (n
14) or 0
as its freezing point (n
4). The re-
sponses of these participants were converted to degrees Fahrenheit.
Anchoring and Adjustment
394 VOL. 12, NO. 5, SEPTEMBER 2001
F(2, 84)
3.89, p
.05. A follow-up contrast showed that partici-
pants provided answers closer to the self-generated anchor (i.e., they
adjusted less) when they were nodding their heads than when they
were shaking their heads, F(1, 42) 6.44, p .05. Participants gave
responses in between those in these two conditions when they were
holding their heads still. Responses to individual items, in raw scores,
are presented in Table 3.
Participants’ head movements also influenced the speed with which
they generated their answers to the questions with self-generated an-
chors, F(2, 84) 5.67, p .05. As predicted, participants answered
more quickly when nodding than when shaking their heads, F(1, 42)
11.76, p .01. The latency of participants’ responses was an interme-
diate value when they were holding their heads still.
We contend that participants adjusted from self-generated anchors
in a serial fashion and that head movements influenced their responses
by altering their willingness to accept values that came initially to
mind. Participants were more willing to accept values that initially
came to mind while nodding their heads, producing less adjustment
and faster reaction times than when they were shaking their heads.
This mechanism differs considerably from the selective-accessibility
mechanism that appears to explain anchoring effects in response to ex-
perimenter-provided anchors, suggesting that different psychological
processes may be operating in these two contexts. Results for the
questions with experimenter-provided anchors are consistent with this
contention: Table 2 shows that participants’ head movements did not
have the same influence on responses to these items.4
Because the strikingly different impact of head movements on re-
sponses to questions with self-generated versus experimenter-provided
anchors is the only evidence of its kind of which we are aware, we
thought it prudent to replicate these results. We thus conducted a close
replication with two changes: (a) We used equal numbers of items
with self-generated and experimenter-provided anchors, and (b) we
counterbalanced the order in which these items were presented. These
changes permitted us to conduct a direct statistical test of the differen-
tial effect of head movements on the two types of questions.
Thirty-two Cornell students participated in a procedure identical to
that of Study 2 except that only 8 questions (4 each with self-gener-
ated and experimenter-provider anchors) were used instead of 16, and
there was no control condition with no head movement. The questions
with self-generated anchors were from Study 2, and the questions with
experimenter-provided anchors were from Jacowitz and Kahneman
(1995)—2 holdovers from Study 2 and 2 new items.
The four items within each anchor type were split into pairs, pro-
ducing two self-generated pairs and two experimenter-provided pairs.
Participants answered one pair of each item type while nodding their
heads, and the other while shaking them. The order in which questions
were presented was counterbalanced and did not influence any of the
results. After each pair, participants evaluated the headphones as part
of the cover story. As in Study 2, participants completed a question-
naire at the end of the session so we could ascertain their knowledge
of the self-generated anchors, and whether they had considered these
anchors when making their estimates.
Results and Discussion
Individual responses were excluded and the data were transformed
in the same manner as in Study 2. Two participants failed to satisfy the
inclusion criteria on at least one item type, leaving 30 participants in
the final analysis.
Participants’ responses to each item were standardized within each
block, and responses were averaged across item type. Participants’
scores were submitted to a 2 (anchor: self-generated vs. experimenter-
provided) 2 (head movement: nodding vs. shaking) repeated mea-
sures analysis of variance. This analysis yielded a marginally signifi-
cant main effect of head movement, F(1, 29) 3.89, p .06,
qualified by the predicted significant interaction, F(1, 29) 9.38, p
.01. As can be seen in Table 4, participants’ answers were more dis-
crepant from a self-generated anchor when they were shaking versus
nodding their heads, paired t(29) 3.61, p .005. Responses to spe-
cific items, in raw scores, are presented in Table 5.
In contrast to the results for the items with self-generated anchors,
head movements did not influence responses to the items with experi-
menter-provided anchors, paired t 1, n.s.5
A similar, although considerably weaker, pattern emerged in an
analysis of participants’ response latencies. As can be seen in Table 4,
participants were somewhat faster to provide answers to the items
with self-generated anchors when they were nodding their heads than
when they were shaking them, paired t(29) 1.52, p .14. Head
movements had no influence on reaction times to questions with ex-
perimenter-provided anchors, paired t 1. The overall interaction be-
tween type of question and amount of time required to generate an
answer, however, was nonsignificant, F(1, 29) 1.77, p .19.
These data replicate those of Study 2 and demonstrate more con-
clusively that self-generated anchors activate a different set of mental
4. Participants’ head movements did not influence their comparative judg-
ments (i.e., whether they believed the true value was higher or lower than the
experimenter-provided anchor), F
1, n.s.
Table 2. Mean standardized answers and reaction times in
Study 2
Head movement
Nodding Still Shaking F(p)a
Self-generated anchors (n
Answer .21 .07 .15 3.89 (.02)
Reaction time .27 .10 .17 5.67 (.005)
Experimenter-provided anchors (n
Answer .16 .25 .07 3.10 (.05)
Reaction time .01 .03 .02 0.03 (n.s.)
aFor items with self-generated anchors, df
2, 84; for items with
experimenter-provided anchors, df
2, 47.
5. As in Study 2, participants’ head movements also did not influence their
comparative judgments, paired t(31)
1.49, p
Nicholas Epley and Thomas Gilovich
VOL. 12, NO. 5, SEPTEMBER 2001 395
operations than experimenter-provided anchors. Head movements in-
fluenced responses when anchors were self-generated but not when
they were provided by the experimenter.
The results of these experiments reestablish the existence of both
anchoring and adjustment in some judgments under uncertainty. When
questions activate self-generated anchors, people adjust from those an-
chors to arrive at final estimates. This process differs considerably
from the processes involved when anchors are provided by an experi-
menter or other external source, demonstrating that there are distinct
anchoring effects produced by different mechanisms. We therefore
second Jacowitz and Kahneman’s (1995) call for a careful taxonomy of
the varieties of anchoring effects in order to advance psychologists’ un-
derstanding of this pervasive element of judgment under uncertainty.
The present experiments have identified the anchor’s source as one
important feature of that taxonomy—a feature that makes it possible
to distinguish those anchoring effects that are produced by a process
of adjustment and those that are not. It is noteworthy in this regard that
a number of phenomena that have been explained through a process of
anchoring and adjustment seem to rely on self-generated anchors sim-
ilar to those that we studied here. These phenomena include trait infer-
ence (Gilbert, in press), interpersonal communication (Keysar & Barr,
in press), comparative ability estimates (Kruger, 1999), and various
egocentric biases (Gilovich et al., 1998, 2000; Keysar & Bly, 1995;
Van Boven, Dunning, & Loewenstein, 2000). Trait inferences begin
with a dispositional attribution that observers generate themselves;
similarly, communication, comparative ability estimates, and the pro-
cesses involved in a host of egocentric judgments begin with a sponta-
neous consideration of one’s own comprehension, skills, or perspective
on the world. Final judgments in these cases are thus likely the prod-
uct of insufficient adjustment from these self-generated anchors. Note
that many of these phenomena are amplified by cognitive-load manip-
ulations designed to hinder any underlying process of adjustment (Gilbert,
in press; Kruger, 1999)—manipulations that have no effect on responses
in the standard anchoring paradigm (Epley & Gilovich, 2000a).
Do adjustments from self-generated anchors tend to be insuffi-
cient? Research on trait inference suggests that although people try to
adjust their impressions to accommodate situational influences, they
adjust too little and are left inferring more about a person’s disposition
than is logically warranted (Gilbert, in press). Research on compara-
tive ability estimates paints a similar picture: Although people try to
adjust for others’ ability level, they adjust too little and are left feeling
systematically above average in domains where absolute skill tends to
be high, such as driving, and below average in domains where it tends
to be low, such as juggling (Kruger, 1999). Results from the control
condition of Study 2 suggest that adjustments in numerical estimates
also tend to be insufficient. Participants in that condition estimated
that George Washington, for example, was elected president in 1779
although he was actually elected in 1788. They also estimated that
vodka freezes at 1.75 F although it actually freezes closer to –20 F.
Indeed, we have reported elsewhere that people tend to systematically
fall short of the actual answer when adjusting from self-generated an-
chors (Epley & Gilovich, 2000b).
This research provides the first compelling evidence that anchoring
effects can be produced by a process that includes adjustment. And al-
though the adjustment process is anything but fully understood, its ex-
istence now seems apparent.
Table 4. Mean standardized answers and reaction times in Study 3
Head movement
t (p)Nodding Shaking
Self-generated anchors (n 30)
Answer .27 .33 3.61 (.001)
Reaction time .16 .07 1.52 (.14)
Experimenter-provided anchors (n 32)
Answer .04 .07 1 (n.s.)
Reaction time .04 .02 1 (n.s.)
Table 3. Mean responses to items with self-generated anchors in Study 2
Question nAnchor
Head movement
Nodding Still Shaking
When was Washington elected president? 37 1776 1777.60 1779.10 1788.10
What is the boiling point of water on Mt. Everest? 32 212 189.31 173.99 141.41
When did the second European explorer, after Columbus, land in
the West Indies? 46 1492 1501.88 1514.36 1548
How many states were in the United States in 1840? 38 50 36.75 30.42 31.64
What is the freezing point of vodka? 38 32 17.36 1.75 9.55
What is the highest recorded body temperature in a human being? 40 98.6 108.34 110.17 107.47
What is the lowest recorded body temperature in a human being? 44 98.6 75.18 83.17 77.65
How many days does it take Mars to orbit the sun?a37 365 127.89 99.36 202.60
What is the gestation period of an African elephant? (months)a45 9 8.50 6.06 5.54
aThe data presented for these items are adjustment scores (the absolute difference between the participant’s answer and his or her reported anchor) because a
number of people adjusted in each direction from the self-generated anchors on these items. Lower numbers indicate a smaller discrepancy between the final
answer and the original anchor (i.e., less adjustment).
Anchoring and Adjustment
396 VOL. 12, NO. 5, SEPTEMBER 2001
Cacioppo, J.T., Priester, J.R., & Berntson, G.G. (1993). Rudimentary determinants of atti-
tudes: II. Arm flexion and extension have differential effects on attitudes. Journal of
Personality and Social Psychology,65, 5–17.
Carlson, B.W. (1990). Anchoring and adjustment in judgments under risk. Journal of Ex-
perimental Psychology: Learning, Memory, and Cognition,16, 665–676.
Cervone, D., & Peake, P.K. (1986). Anchoring, efficacy, and action: The influence of judg-
mental heuristics on self-efficacy judgments and behavior. Journal of Personality
and Social Psychology,50, 492–501.
Chapman, G.B., & Johnson, E.J. (1994). The limits of anchoring. Journal of Behavioral
Decision Making,7, 223–242.
Chapman, G.B., & Johnson, E.J. (in press). Incorporating the irrelevant: Anchors in judg-
ments of belief and value. In T. Gilovich, D. Griffin, & D. Kahneman (Eds.), Heu-
ristics and biases: The psychology of intuitive judgment. Cambridge, England:
Cambridge University Press.
Crocker, J. (1982). Biased questions in judgment of covariation studies. Personality and
Social Psychology Bulletin,8, 214–220.
Epley, N., & Gilovich, T. (2000a). Anchoring and effortful adjustment from self-generated
anchors. Unpublished manuscript, Cornell University, Ithaca, NY.
Epley, N., & Gilovich, T. (2000b). Close but not quite: Insufficient adjustment from self-
generated anchors. Unpublished manuscript, Cornell University, Ithaca, NY.
Fischhoff, B., & Beyth, R. (1975). “I knew it would happen”—remembered probabilities
of once-future things. Organizational Behavior and Human Performance,13, 1–16.
Forster, J., & Strack, F. (1996). Influence of overt head movements on memory for va-
lenced words: A case of conceptual-motor compatibility. Journal of Personality and
Social Psychology,71, 421–430.
Forster, J., & Strack, F. (1997). Motor actions in retrieval of valenced information: A mo-
tor congruence effect. Perceptual and Motor Skills,85, 1419–1427.
Gilbert, D.T. (1989). Thinking lightly about others: Automatic components of the social
inference process. In J.S. Uleman & J.A. Bargh (Eds.), Unintended thought (pp.
189–211). New York: Guilford.
Gilbert, D.T. (in press). Inferential correction. In T. Gilovich, D. Griffin, & D. Kahneman
(Eds.), Heuristics and biases. Cambridge, England: Cambridge University Press.
Gilovich, T., Medvec, V.H., & Savitsky, K. (2000). The spotlight effect in social judgment:
An egocentric bias in estimates of the salience of one’s own actions and appearance.
Journal of Personality and Social Psychology,78, 211–222.
Gilovich, T., Savitsky, K., & Medvec, V.H. (1998). The illusion of transparency: Biased as-
sessments of others’ ability to read our emotional states. Journal of Personality and
Social Psychology,75, 332–346.
Hawkins, S.A., & Hastie, R. (1991). Hindsight: Biased judgments of past events after the
outcomes are known. Psychological Bulletin,107, 311–327.
Jacowitz, K.E., & Kahneman, D. (1995). Measures of anchoring in estimation tasks. Per-
sonality and Social Psychology Bulletin,21, 1161–1167.
Keysar, B., & Barr, D.J. (in press). Self-anchoring in conversation: Why language users
don’t do what they “should.” In T. Gilovich, D. Griffin, & D. Kahneman (Eds.),
Heuristics and biases. Cambridge, England: Cambridge University Press.
Keysar, B., & Bly, B. (1995). Intuitions of the transparency of idioms: Can one keep a se-
cret by spilling the beans? Journal of Memory and Language,34, 89–109.
Kruger, J. (1999). Lake Wobegon be gone! The “below-average effect” and the egocentric
nature of comparative ability judgments. Journal of Personality and Social Psychol-
ogy,77, 221–232.
Lichtenstein, S., & Slovic, P. (1971). Reversals of preference between bids and choices in
gambling decisions. Journal of Experimental Psychology,89, 46–55.
Martin, L.L., Harlow, T.F., & Strack, F. (1992). The role of bodily sensations in the evalu-
ation of social events. Personality and Social Psychology Bulletin,18, 412–419.
Mussweiler, T., & Strack, F. (1999). Hypothesis-consistent testing and semantic priming in
the anchoring paradigm: A selective accessibility model. Journal of Experimental
Social Psychology,35, 136–164.
Mussweiler, T., & Strack, F. (2000). The use of category and exemplar knowledge in the solu-
tion of anchoring tasks. Journal of Personality and Social Psychology,78, 1038–1052.
Mussweiler, T., & Strack, F. (in press). “Considering the impossible”: Explaining the ef-
fects of implausible anchors. Social Cognition.
Nisbett, R.E., & Wilson, T.D. (1977). Telling more than we can know: Verbal reports on
mental processes. Psychological Review,84, 231–259.
Plous, S. (1989). Thinking the unthinkable: The effect of anchoring on likelihood esti-
mates of nuclear war. Journal of Applied Social Psychology,19, 67–91.
Priester, J.R., Cacioppo, J.T., & Petty, R.E. (1996). The influence of motor processes on at-
titudes toward novel versus familiar semantic stimuli. Personality and Social Psy-
chology Bulletin,22, 442–447.
Schkade, D.A., & Johnson, E.J. (1989). Cognitive processes in preference reversals. Orga-
nizational Behavior and Human Decision Processes,44, 203–231.
Snyder, M., & Swann, W.B. (1978). Hypothesis-testing processes in social interaction.
Journal of Personality and Social Psychology,36, 1202–1212.
Strack, F., & Mussweiler, T. (1997). Explaining the enigmatic anchoring effect: Mechanisms
of selective accessibility. Journal of Personality and Social Psychology,73, 437–446.
Switzer, F., & Sniezek, J.A. (1991). Judgmental processes in motivation: Anchoring and
adjustment effects on judgment and behavior. Organizational Behavior and Human
Decision Processes,49, 208–229.
Trope, Y., & Bassok, M. (1982). Confirmatory and diagnosing strategies in social informa-
tion gathering. Journal of Personality and Social Psychology,43, 22–34.
Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases.
Science,185, 1124–1130.
Van Boven, L., Dunning, D., & Loewenstein, G. (2000). Egocentric empathy gaps be-
tween owners and buyers: Misperceptions of the endowment effect. Journal of Per-
sonality and Social Psychology,79, 66–76.
Wells, G.L., & Petty, R.E. (1980). The effects of overt head movements on persuasion:
Compatibility and incompatibility of responses. Basic and Applied Social Psychol-
ogy,1, 219–230.
Wright, W.F., & Anderson, U. (1989). Effects of situation familiarity and incentives on use
of the anchoring and adjustment heuristic for probability assessment. Organiza-
tional Behavior and Human Decision Processes,44, 68–82.
(RECEIVED 9/25/00; ACCEPTED 12/15/00)
Acknowledgments—This research was supported by Research Grant
SBR9809262 from the National Science Foundation. We would like to thank
Sabiha Barot, Noah Goldstein, Thalia Goldstein, Ellyn Poltrock, Brett Robinson,
and Kevin Van Aelst for their help collecting data, and Dennis Regan and Leaf
Van Boven for helpful comments throughout this research.
Table 5. Mean answers to items with self-generated anchors in Study 3
Question nAnchor
Head movement
Nodding Shaking
When was Washington elected president? 28 1776 1783.50 1788.25
When did the second European explorer, after Columbus, land in the West Indies? 30 1492 1508.72 1534.42
What is the boiling point on Mt. Everest? 21 212 192.50 176.90
What is the freezing point of vodka? 28 32 12.47 19.09
... Initial anchor value tends to influence the direction of the individuals' final judgment or estimation (Jacowitz & Kahneman, 1995). Externally provided anchor plays a significant role than a selfgenerated anchor and has more weight and validity and therefore is assumed to be closely related to the correct answer (Epley & Gilovich, 2001). EWOM is considered to be an important source of information as it is created by other customers and the anchoring effect would fit this scenario as well. ...
Being authentic in advertising and using electronic word of mouth (eWOM) to satisfy customers pose significant challenges to managers. Drawing from prior literature, social influence theory, and the anchoring effect this study offers a framework to develop and apply authenticity in advertising and to use eWOM as marketing tools to satisfy customers. Consistent with prior literature, this framework identifies four dimensions of authenticity in advertising when implementing it as a strategy in marketing: brand essence, brand heritage, realistic plot, and message credibility. Firms may choose to attach importance to any of the dimensions of authenticity. Depending on the firm's strategy, managers can use this framework to aid firms to examine and use authenticity in advertising. The authors formulate hypotheses to suggest that depending on the firm's strategy, the effect of each of the four dimensions of authenticity on positive eWOM can be managed differently. Since few studies have focused on the impact of positive eWOM on customer satisfaction, this study addresses the issue by also considering the moderating effect of perceived trustworthiness. This study will help managers to get a better idea regarding how to execute authenticity in advertising and to use positive eWOM to satisfy customers.
... First, anchoring bias proves to be robust across a variety of experimental manipulations and contextual factors (Furnham & Boo, 2011). Even if anchors are self-generated, people do not adjust their estimates sufficiently (Epley & Gilovich, 2001). On top of that, increased understanding of the problem at hand does not discard anchoring bias. ...
Full-text available
The rise of behavioral public administration demonstrated that we can understand and change decision-making by using insights about heuristics. Heuristics are mental shortcuts that reduce complex tasks to simpler ones. Whereas earlier studies mainly focused on interventions such as nudges, scholars are now broadening their scope to include debiasing, and psychological theories beyond heuristics. Scholars are moreover shifting their attention away from citizen-focused interventions to public sector worker-oriented interventions, i.e. the very people who are expected to nudge society. This dissertation seeks to explore how behavioral sciences can facilitate understanding and support decision-making across the public sector. We present four studies that investigate a range of behavioral theories, practices, issues and public sector workers. This dissertation shows that when handling heuristics in the public sector, we need to take into account the institutional and situational settings, as well as differences between public sector workers. The results of this dissertation can be used by practitioners and academics to understand and support decision-making in public sector contexts.
... This heuristic describes how people often start with a specific piece of information (the anchor) and adjust from there to come up with a final answer. However, research has demonstrated that there is a systematic bias to make insufficient adjustments from the anchor (Epley & Gilovich, 2001). The anchor can be relevant, for example, the initial offer in the process of a house sale or wage negotiation. ...
... No difference between anchor types EP + SA and SG + IA Bahník (2021): the selective accessibility model is problematic • Epley and Gilovich (2001, 2004, 2005, 2006: Self-generated anchors are moderated by need for cognition, forewarning, incentives, cognitive load, head movement, arm flexion, and alcohol consumption REPLICATING EPLEY AND GILOVICH 5 ...
Full-text available
Anchoring, the assimilation of numerical estimates toward previously considered numbers, has generally been separated into anchoring from self-generated anchors (e.g., people first thinking of 9 months when asked for the gestation period of an animal) and experimenter-provided anchors (e.g., experimenters letting participants spin fortune wheels). For some time, the two types of anchoring were believed to be explained by two different theoretical accounts. However, later research showed crossover between the accounts. What now remains are contradictions between past and recent findings, specifically, which moderators affect which type of anchoring. We conducted three replications (Ntotal = 653) of seminal studies on the distinction between self-generated and experimenter-provided anchoring effects where we investigated the moderators need for cognition, cognitive load, and forewarning. We found no evidence that either type of anchoring is moderated by any of the moderators. In line with recent replication efforts, we found that anchoring effects were robust, but the findings on moderators of anchoring effects should be treated with caution.
This study details the results of an experimental intervention designed to address the issue of price anchoring in the choice experiment framework. The intervention, which informs respondents of the tendency to anchor choices on potentially arbitrary pieces of information, is applied to a choice experiment used to examine consumers’ willingness to pay for local and/or organic tomatoes in Northern New England and develops three primary contributions. First, evidence from this study shows that anchoring effects are present. Second, providing information to consumers plays a mitigating role on these effects; price anchoring changes willing to pay estimates between 44% and 51% and exposure to anchoring-specific cheap talk is associated with a reduction in these anchoring effects between 60% and 80%. These results are explained through decreases in price sensitivity induced by increasing the mean price vector and subsequent increases in price sensitivity due to the information intervention. Finally, this study reveals that consumers are willing to pay a substantial price premium for locally grown tomatoes, from $0.96 to $1.12 per pound, offering some guidance for policy regarding growing practice and farm land use as regional coalitions support local agriculture expansion in the Northeast.
Purpose The objective of this paper is to explore how anchoring affects the dynamics of investor decision-making with regard to mutual funds and how this bias differs amongst gender and level of financial knowledge. Design/methodology/approach An experimental research design was adopted to uncover the relationship between the variables under study; this involved the use of a questionnaire with an embedded experiment. Data obtained from the study were analysed using Pearson's chi-square test and two-way analysis of variance. Findings The findings show that, overall, investors were prone to be significantly influenced by the anchoring bias. The study finds a strong, albeit not significant, association between participants' susceptibility to anchor and both gender and the level of financial knowledge of participants. Females were observed to be more likely to anchor than their male counterparts. Also, a higher level of financial knowledge did not help to reduce the possibility of anchoring; it rather increased it. Research limitations/implications The findings of the study cannot be interpreted as suggesting causality as the study only tests for association between variables and not causality. Additionally, external validity cannot be fully established as a result of the quasi-experiment approach used. Practical implications The study adds to the body of knowledge on the influences of behavioural biases in the sub-region to make investors aware of their biases in order to minimise the influence of these biases on their investment decisions. Originality/value This study differs from earlier studies in that it analyses the presence of anchoring as influenced by a completely different set of variables (expertise and gender) and also does it within the context of an African country where there remains a paucity of research on behavioural finance.
This research investigates how up- and down-selling promotions affect customers’ cross-brand purchasing and churn behavior at a multi-brand retailer. We employ a hidden Markov model that accounts for the dynamics of customers’ latent brand preferences and attrition and captures the resulting purchase behavior in response to promotional offers. Using data on coupon promotions and purchase transactions from an online retailer, we find that coupons for a higher-end brand increase customers’ likelihood of purchasing the brand. While this suggests that the retailer can increase its short-term revenues by sending coupons for the higher-end brand to customers of the lower-end brand, we find that customers up-sold via coupons are more likely to switch back to the lower-end brand, in comparison to other customers of the higher-end brand, limiting the positive effect of up-selling promotions in the long term. Moreover, lower-end brand customers’ promotion-induced brand switching leads to their increased attrition from the retailer, which negatively affects long-term purchase behavior and revenues. In contrast, when triggered by down-selling coupons, customers’ brand switching does not impact their attrition. Based on these findings, we demonstrate how our model-based approach can assist marketers’ multi-brand couponing decisions.
Full-text available
Cognitive scientists have revealed systematic errors in human reasoning. There is disagreement about what these errors indicate about human rationality, but one upshot seems clear: human reasoning does not seem to fit traditional views of human rationality. This concern about rationality has made its way through various fields and has recently caught the attention of philosophers. The concern is that if philosophers are prone to systematic errors in reasoning, then the integrity of philosophy would be threatened. In this paper, I present some of the more famous work in cognitive science that has marshaled this concern. Then I present reasons to think that those with training in philosophy will be less prone to certain systematic errors in reasoning. The suggestion is that if philosophers could be shown to be less prone to such errors, then the worries about the integrity of philosophy could be constrained. Then I present evidence that, according to performance on the CRT (Frederick 2005), those who have benefited from training and selection in philosophy are indeed less prone to one kind of systematic error: irrationally arbitrating between intuitive and reflective responses. Nonetheless, philosophers are not entirely immune to this systematic error, and their proclivity for this error is statistically related to their responses to a variety of philosophical questions. So, while the evidence herein puts constraints on the worries about the integrity of philosophy, it by no means eliminates these worries. The conclusion, then, is that the present evidence offers prima facie reasons to ascribe a mitigated privilege to philosophers' ability to rationally arbitrate between intuitive and reflective responses.
Subjective expectation data on education has been increasingly used by social scientists to better understand current investments in human capital. Despite its recognised value by scholars, there is little evidence about how the elicitation of such data might be sensitive to questionnaire design. Using a 2x2 between-subjects experimental design, we analyse how sensitive the elicitation of subjective expectation data on educational outcomes is to anchors. Our study provides causal evidence on whether collecting data on parental education before the elicitation of parental expectations on their children’s educational outcomes anchors the elicitation of the latter; and whether parental expectations on their older offsprings anchors their expectations on their younger children. We find that mothers (main respondents) who have been exposed to the anchored treatments report more pessimistic parental expectations. When splitting our sample into low and high educated mothers, we find that low educated mothers who have been allocated to anchored treatments are more likely to report lower levels of education than those in the non-anchored treatment. Anchored treatments also increase non-response in both high and low educated mothers, however, the effect is larger on the former. When assessing the accuracy of expectations to predict educational outcomes, we observe that anchored expectations have higher predictive power. Our findings inform to what extent the collection of subjective expectations data is subject to anchoring and which type of elicitation (anchored or non-anchored) should be considered according to the main purpose of the elicitation (i.e., item response vs prediction).
Full-text available
Results of four studies support the notion that anchoring effects are mediated by mechanisms of hypothesis-consistent testing and semantic priming. According to the suggested Selective Accessibility Model, judges use a hypothesis-consistent test strategy to solve a comparative anchoring task. Applying this strategy selectively increases the accessibility of anchor-consistent knowledge which is then used to generate the subsequent absolute judgment. Studies 1 and 2 demonstrate that absolute estimates depend on the hypothesis implied in the comparative task, suggesting that a hypothesis-testing strategy is used to solve this task. Study 3 shows that limiting the amount of knowledge generated for the comparative task retards absolute judgments. This suggests that knowledge rendered easily accessible in the comparative judgment is used for the subsequent absolute judgment. Finally, Study 4 suggests that self-generation of knowledge contributes to the robustness of the effect, thus resolving the seeming inconsistency that anchoring effects are at the same time remarkably robust and mediated by typically fragile semantic priming mechanisms.
Full-text available
Recent research has suggested that the motor processes elicited by upper arm isometric flexion and extension can subtly influence attitudes. Specifically, ideographs evaluatively categorized while performing isometric upper arm flexion were subsequently preferred to ideographs evaluatively categorized while performing isometric upper arm extension. The present research sought to replicate this attitudinal effect with semantic stimuli and to examine a theoretical boundary condition posited by the elaboration likelihood model (ELM). Subjects categorized either neutral words or pronounceable nonwords while adopting the isometric positions. Results indicated that motor processes influenced subsequent attitudes toward stimuli with few associations (i.e., nonwords) more than toward stimuli with many associations in memory (i.e., familiar words). These results are consistent with a growing literature on the possible influence of nondeclarative (e.g., procedural) knowledge on attitudes.
Full-text available
Research on judgmental anchoring - the assimilation of a numeric estimate towards a previously considered standard - has demonstrated that implausible anchors pro- duce large effects. We propose an insufficient adjustment plus selective accessibil- ity account for these effects. Specifically, judges may adjust from an implausible anchor until a plausible value for the target is reached and may then test the hypoth- esis that the target's extension is similar to this value. If this is indeed the case, then differentially extreme implausible anchors should produce similar absolute esti- mates, because adjustment from any implausible anchor should terminate at the same value. Results of two studies are consistent with this prediction. They show that implausible anchors that differ extremely produce similar absolute estimates. The implications of these findings for alternative models of anchoring are dis- cussed. Human judgment under uncertainty is often influenced by salient judgmental anchors. In what is probably the best known demonstra- tion of such anchoring effects (Tversky & Kahneman, 1974), partici- pants first received a comparative judgment task in which they were asked whether the percentage of African nations in the UN is higher or lower than an arbitrary number (the anchor) that had been deter- mined by spinning a wheel of fortune (i.e., 65% or 10%). In the subse- quent absolute judgment task, participants were asked to give their
Evidence is reviewed which suggests that there may be little or no direct introspective access to higher order cognitive processes. Subjects are sometimes (a) unaware of the existence of a stimulus that importantly influenced a response, (b) unaware of the existence of the response, and (c) unaware that the stimulus has affected the response. It is proposed that when people attempt to report on their cognitive processes, that is, on the processes mediating the effects of a stimulus on a response, they do not do so on the basis of any true introspection. Instead, their reports are based on a priori, implicit causal theories, or judgments about the extent to which a particular stimulus is a plausible cause of a given response. This suggests that though people may not be able to observe directly their cognitive processes, they will sometimes be able to report accurately about them. Accurate reports will occur when influential stimuli are salient and are plausible causes of the responses they produce, and will not occur when stimuli are not salient or are not plausible causes.
Many decisions are based on beliefs concerning the likelihood of uncertain events such as the outcome of an election, the guilt of a defendant, or the future value of the dollar. Occasionally, beliefs concerning uncertain events are expressed in numerical form as odds or subjective probabilities. In general, the heuristics are quite useful, but sometimes they lead to severe and systematic errors. The subjective assessment of probability resembles the subjective assessment of physical quantities such as distance or size. These judgments are all based on data of limited validity, which are processed according to heuristic rules. However, the reliance on this rule leads to systematic errors in the estimation of distance. This chapter describes three heuristics that are employed in making judgments under uncertainty. The first is representativeness, which is usually employed when people are asked to judge the probability that an object or event belongs to a class or event. The second is the availability of instances or scenarios, which is often employed when people are asked to assess the frequency of a class or the plausibility of a particular development, and the third is adjustment from an anchor, which is usually employed in numerical prediction when a relevant value is available.
According to the anchoring and adjustment heuristic, an individual anchors an initial judgment of a stimulus with some of its features and then adjusts that initial judgment to reflect the remaining features. This article examines anchoring and adjustment in the assignment of minimum selling prices (MSPs) to three-outcome gambles. The results of Experiment 1 indicated that people were most heavily influenced by the intermediate outcome in a gamble. On the basis of these results, the following framework was proposed: To forecast what is likely to happen in a play of a gamble, people partition the outcomes into sets based on the similarity of the outcomes. They then form an initial judgment based on the outcomes in the more probable set, which they adjust to reflect the remaining outcomes. Two more experiments show that when one outcome has a probability greater than.50, people do not focus on the intermediate outcome but anchor their judgment with the high-probability outcome.
The authors describe a method for the quantitative study of anchoring effects in estimation tasks. A calibration group provides estimates of a set of uncertain quantities. Subjects in the anchored condition first judge whether a specified number (the anchor) is higher or lower than the true value before estimating each quantity. The anchors are set at predetermined percentiles of the distribution of estimates in the calibration group (15th and 85th percentiles in this study). This procedure permits the transformation of anchored estimates into percentiles in the calibration group, allows pooling of results across problems, and provides a natural measure of the size of the effect. The authors illustrate the method by a demonstration that the initial judgment of the anchor is susceptible to an anchoring-like bias and by an analysis of the relation between anchoring and subjective confidence.