6. The Sin of Bias.

The Sin of Bias.

IN GEORGE ORWELL'S chilling novel of life in a totalitarian political system, 1984, the ruling party achieved psychological mastery over its subject by willfully altering the past. "Who controls the past," ran the party slogan,
"controls the future: who controls the present controls the pase' The government's Ministry of Truth tried to alter the written historical record and
even to manipulate the actual experience of remembering:

Past events, it is argued, have no objective existence, but survive only in written records and in human memories. The past is whatever the records and the memories agree upon ... control of the past depends
above all on the training of memory. To make sure that all written records agree with the orthodoxy of the moment is merely a mechanical
act. But it is also necessary to remember that events happened in the desired manner. And if it is necessary to rearrange one's memories or to
tamper with written records, then it is necessary to forget that one has done so. The trick of doing this can be learned like any other mental
technique.

Totalitarian societies like the one envisioned by Orwell have declined
since the collapse of the eastern European communist regimes. But forces
that in some sense resemble the Ministry of Truth continue to operate in
individual minds: our memories of the past are often rescripted to fit with
our present views and needs. The sin of bias refers to distorting influences
of our present knowledge, beliefs, and feelings on new experiences or our
later memories of them. In the stifling psychological climate of 1984, the
Ministry of Truth used memory as a pawn in the service of party rule.
Much in the same manner, biases in remembering past experiences reveal
how memory can serve as a pawn for the ruling masters of our cognitive
systems.

Five major types of biases illustrate the ways in which memory serves
its masters. Consistency and change biases show how our theories about
ourselves can lead us to reconstruct the past as overly similar to, or different
from, the present. Hindsight biases reveal that recollections of past events
are filtered by current knowledge. Egocentric biases illustrate the powerful
role of the self in orchestrating perceptions and memories of reality. And
stereotypical biases demonstrate how generic memories shape interpretation of the world, even when we are unaware of their existence or influence.

THE WAY WE WERE DEPENDS ON THE WAY WE ARE.

When Ross Perot unexpectedly announced his withdrawal from the presidential race on July 16, 1992, he dealt a cruel blow to his fervent supporters.
Perot was widely reviled in the press - Newsweek ran a cover story on him
titled "The Quitter" - and his allies experienced a complex mixture of
sadness, anger, and hope that he might reconsider the decision. When he
reentered the campaign in early October, those who had supported him reacted in different ways. Loyalists never wavered from Perot and renewed
their efforts on his behalf. Returning supporters had initially switched to
another candidate but quickly came back. Deserters abandoned Perot as
soon as he left the race and never returned.

A few days after Perot quit in July, the University of California at
Irvine psychologist Linda Levine asked his supporters how they felt; she
then probed their memories again after the election in November. Loyalists, returning supporters, and deserters all accurately recalled, at least to
some degree, the sadness, anger, and hope they had felt when Perot made his stunning July announcement. But they also rewrote their memories to
be consistent with how they felt in November. After the election, loyalists
underestimated how sad they felt when Perot quit. Returning supporters
recalled feeling less angry in July than they actually said they were at the
time. And deserters recalled being less hopeful than they actually were.

This consistency bias has turned up in several different contexts. Recalling past experiences of pain, for instance, is powerfully influenced by
current pain level. When patients afflicted by chronic pain are experiencing
high levels of pain in the present, they are biased to recall similarly high levels of pain in the past; when present pain isn't so bad, past pain experiences
seem more benign, too. Attitudes toward political and social issues also reflect consistency bias. People whose views on political issues have changed
over time often recall incorrectly past attitudes as highly similar to present
ones. In fact, memories of past political views are sometimes more closely
related to present views than to what people actually believed in the past. In
one study, high school students stated their opinions on school busing, and
then heard arguments for or against busing. Despite changing their views
in line with the arguments they heard, the students mistakenly recalled that
they had always held the views they expressed after hearing the pro or con
argument.

To appreciate why people are so prone to consistency biases, try to recall your views on capital punishment five years ago. Can you specifically
recall what you believed in those days? The Canadian social psychologist
Michael Ross has observed that people often do not have clear memories of
exactly what they believed or felt in the past, and instead infer past beliefs,
attitudes, and feelings from their current states. Unless there is good reason
to believe that your views on capital punishment have changed, you are
likely to assess your present opinion and assume you felt the same way five
years ago. Invoking what Ross calls an "implicit theory of stability" will lead
to accurate recall if your views haven't changed over time, but will produce
a consistency bias if they have.

People don't always invoke a theory of stability, however; sometimes
we believe that we have, or should have, changed over time. Self-help programs may exploit such feelings. Once people invest time and energy in a
program that is supposed to help them change - lose weight, prepare for
college entrance exams, or exercise more - they may exaggerate the degree
of change they've actually experienced. Students who completed a program
purported to enhance their study skills remembered their initial level of
skill as being lower than they said it was before beginning the program,
whereas students who were on a waiting list for the program showed no
change bias.

Change bias also influences how women recall their emotional states
during menstruation. Surveys indicate that women generally believe that
they are likely to become highly irritable and depressed during periods.
Studies of women during menstruation clearly show heightened incidence
of such physical symptoms as backaches, headaches, and abdominal pains
but there's little evidence for greater depression or related mood changes. Physical discomfort may lead women to theorize that menstruation results in negative moods and related kinds of psychological distress.

In a study from Michael Ross's group, women who were menstruating
reported more physical symptoms compared to when they were not, and
showed little change in self-reported mood or personality measures. Yet
during menstruation, these women recalled intermenstrual emotional
states as more positive than they actually were, supporting their theories
that menstruation produces bad moods. Such theories can also inflate recall of negative menstrual symptoms: the more a woman believes that she
experiences bad moods during menstruation, the more she shows exaggerated recall of such symptoms after her period has concluded.
The effects of consistency and change bias are perhaps nowhere more
evident than in recollections of dose personal relationships. Recall the
1970S Barbra Streisand tune "The Way We Were":

Memories.
May be beautiful, and yet,
What's too painful to remember.
We simply choose to forget;
For it's the laughter.
We will remember.
Whenever we remember.
The way we were.

As implied by the song, as well as by the evidence and ideas considered
so far, it is difficult to separate recall of "the way we were" from current appraisals of "the way we are." Consistency biases often color couples' retrospective assessments of how they once felt, with the present state of the
relationship dictating memory for how things used to be. Consider, for instance, dating college students who were asked, in separate sessions conducted two months apart, to evaluate themselves and their dating partners
on such traits as honesty, kindness, intelligence, and also according to how
much they like and love their partners. During the second session, the couples also recalled earlier evaluations. The students whose evaluations of
their partners became more negative over time recalled their initial impressions as more negative than they actually were. Students who reported
liking or loving a partner more in the second session than in the first also
recalled having felt more love or liking in the past. Memories of past impressions and feelings were filtered through, and made consistent with,
partners' current impressions and feelings.

Consistency biases are prevalent in both married and dating couples.

Consider the following questions in relation to your own partner: How attached do you feel? How happy are you in your relationship? How often
does your partner get on your nerves? How much do you love him or her?
Then try to answer the same questions, instead focusing on how you felt a
year ago. Married and dating couples who were asked similar questions
twice, over a period of eight months or four years, often remembered correctly that they had given similar ratings on the two occasions. But those
men and women whose feelings had changed over time tended to mistakenly remember that they had always felt the same way. Trying to remember
what they felt four years earlier, four out of five people whose feelings remained stable showed accurate recall, but only one in five of those whose
feelings had changed recalled accurately "the way they were." Results were
even more dramatic when couples recalled how they had felt eight months
earlier: 89 percent of women and 85 percent of men whose feelings remained stable accurately remembered their initial impressions, but only 22
percent of women and 15 percent of men whose feelings had changed
showed accurate recall. The couples seemed to be saying "what I feel now is
what I've always felt" - regardless of whether they had or not.

These kinds of biases can sometimes accentuate troubles that some
married couples experience during their first few years together. Once the
"honeymoon" is over, many couples experience a sharp drop-off in levels
of satisfaction with their marriages. Difficulties in the present are hard
enough to address during the early years of marriage, but consistency biases can make matters worse by coloring the past with the unpleasant tones
of the present. Consider a study that followed nearly four hundred Michigan couples through the first years of marriage. In those couples who expressed growing unhappiness over the four years of the study, men mistakenly recalled the beginnings of their marriages as negative even though
they said they were happy at the time. "Such biases can lead to a dangerous
downward spiral;' noted the researchers who conducted the study. "The
worse your current view of your partner is, the worse your memories are,
which only further confirms your negative attitudes:'

Though consistency biases are potent forces in shaping relationship
memories, change biases can also occur - and sometimes in a positive direction. Remember the popular late-1960s song with the line "I love you
more today than yesterday"? People would no doubt like to believe that
their romantic attachments grow stronger over time. When dating couples

were asked, once annually, to assess the present quality of their relationships and to recall how they felt in past years, their recollections embodied
the same sentiments as the song line. Couples who stayed together for the
four years recalled that the strength of their love had grown since they last
reported on it. Yet analysis of their actual ratings at the time failed to show
any increases in reported love and attachment. Objectively, the couples did
not love each other more today than yesterday. But through the subjective
lenses of memory, they did.

This pattern differs from the consistency biases seen in other dating
and married couples, instead revealing a kind of improvement bias. The
couples mistakenly remembered the past as less positive than it actually
was, making the present seem rosier by comparison. Consistency and
change biases can each occur at different points in a relationship, with the
predominant bias at a particular time depending on the nature and stage of
the relationship. Benjamin Kearney from the University of Florida and
Robert Coombs from the University of California at Los Angeles analyzed a
twenty-year longitudinal investigation of wives' feelings about their marriages. The study was initiated in 1969, when the women were in their midtwenties. The scientists separately considered the first ten years of the
women's marriages, when couples were making the transition to parenthood, and the second ten years, as they entered a period of personal and
economic stability. On each occasion, wives answered questions, some very
general (How happy are you with your marriagd), others more specific
(How many interests do you and your husband share?).

When reflecting back on the first ten years of their marriages, wives
showed a change bias: they remembered their initial assessments as worse
than they actually were. The bias made their present feelings seem an improvement by comparison, even though the wives actually felt more negatively ten years into their marriage than they had at the beginning. When
they had been married for twenty years and reflected back on their second
ten years of marriage, the women now showed a consistency bias: they mistakenly recalled that feelings from ten years earlier were similar to their
present ones. In reality, however, they felt more negatively after twenty
years of marriage than after ten. Both types of bias helped women cope
with their marriages. The more women's recollections were biased toward
improvement at the ten-year mark, the happier they were with their marriages at the twenty-year mark. By the twenty-year mark, wives who were
most satisfied with their marriages showed the least memory bias, whereas
those who were least satisfied showed the most bias - perhaps reflecting
ongoing attempts to cope with an unhappy present by distorting the past.
Memories of "the way we were" are not only influenced by, but also contribute to, "the way we are."

Consistency and change biases may help to reduce what social psychologists call "cognitive dissonance" - the psychological discomfort that
results from conflicting thoughts and feelings. People will go to great
lengths to reduce cognitive dissonance. A heavy drinker who reads the latest health statistics highlighting the dangers of excessive alcohol intake
might try to reduce dissonance by convincing himself that he is only a light
social drinker or by disparaging the statistics. Likewise, an unhappily married woman who believes that her marriage should be successful may reduce cognitive dissonance by distorting the past with consistency or change
biases that make the present seem more bearable.

Dissonance reduction can occur even when people don't recall the
event that is responsible for the dissonance. Consider the following scenario. You visit an art gallery and fall in love with two prints by the same
artist, but have only enough money to purchase one. After almost deciding
on one and then the other, you finally make your choice, but as you leave
with your new purchase you still feel conflicted about passing over the remaining print. By the next day, however, you realize that you like the print
you purchased quite a bit more than the one you passed up, and the dissonance created by the difficult decision dissipates.

Studies have shown that just this sort of dissonance reduction occurs
when people are forced to decide between two art prints that they previously indicated they liked equally: after making the choice they claim to like
the chosen print more and the bypassed print less than they had earlier. In
a study led by the social psychologists Matthew Lieberman and Kevin
Ochsner, we found that amnesic patients also reduced dissonance created
by choosing between two art prints they liked equally by later inflating how
much they liked the chosen relative to the shunned print. But the amnesic
patients had no conscious memory for making the choice that produced
dissonance in the first place! These findings suggest that a variety of dissonance-reducing operations, including consistency and change biases, occur
even when people have limited awareness of the source of the conflicts they
are trying to manage.

I KNEW IT ALL ALONG.

When the Boston Red Sox beat the Cleveland Indians in the deciding game
of their playoff series in October 1999, Boston sports fans relished the prospects of taking on the world champion New York Yankees in the American
League championship series. Euphoric callers to sports radio talk shows
enumerated reason after reason why the long-suffering Red Sox had an excellent chance to dethrone the mighty - and hated - Yankees. The Red
Sox had built up tremendous momentum in their come-from-behind win
against Cleveland; no team could hit their dominating pitcher, Pedro Martinez; and in a short baseball series, anything can happen.

After the Red Sox lost the series, the talk show callers reasoned very
differently. I never thought the Red Sox had a chance, caller after caller
grimly stated. I was sure they didn't have enough hitting to compete, recollected some; I always felt their bullpen was too weak, remembered others. I
knew that the Yankees were too good, even diehard Red Sox fans conceded.

The memories of the talk show callers seemed to be powerfully influenced by the outcome of the playoff series: with the benefit of hindsight,
the fans felt that they knew all along that the Red Sox were doomed to
lose. Although it's difficult to draw firm conclusions based on an unscientific sampling of opinions on radio show talks (perhaps optimistic callers
phoned in before the playoffs and the pessimists held off until later), controlled studies of other sports fans back up this interpretation of what the
Red Sox callers said. Followers of the Northwestern University football
team assessed the team's prospects of winning, either before or after home
games played during the fall 1995 season against Wisconsin, Penn State, and
Iowa. Northwestern, which enjoyed a highly successful season in 1995, won
all three games. Fans who were asked after each game to recall what they
had thought before the game gave Northwestern a much greater chance of
winning than fans who rated the team's chances before the game.

Sports fans aren't the only ones who "knew it all along:' Consider another public event that people expressed strong opinions about: the jury
decision in O. J. Simpson's criminal trial. Can you recall how likely you
thought the jury was to convict O.J.? Students were asked to estimate the
likelihood that the jury would convict 0.1. two hours before the jurors returned a not-guilty verdict, and again two days later, after the students
knew the verdict. They rated the likelihood of conviction as lower after the
jury made its decision than before.

Judgments about sports events and the O.J. trial illustrate a familiar
occurrence in everyday life: once we learn the outcome of an event, we feel
as though we always knew what would happen. Called hindsight bias by
psychologists, this tendency to see an outcome as inevitable in retrospect is
a dose cousin of consistency bias: we reconstruct the past to make it consistent with what we know in the present.

Hindsight bias seems particularly common around the time of political elections, with various pundits rushing to explain why the outcome of a
particular race could hardly have gone otherwise. But did they see things so
clearly before the votes were counted? On the day before the 1980 presidential election, students were asked to predict the outcome. Others were asked
the day after to indicate what they would have predicted before the Tuesday
election. Those who were asked on Wednesday "predicted" a higher percentage of the vote for Reagan and lower percentages for Jimmy Carter and
independent candidate John Anderson than those who had been asked on
Monday.

Hindsight bias is especially pronounced when people come up with
after-the-fact explanations that specify a deterministic cause of the outcome. Consider, for instance, people who judged alternate outcomes of a
nineteenth-century war between the British and the Gurkas of Nepal. In
the foresight condition, people read about the incident and judged the likelihood of various outcomes. In the hindsight condition, people were told
the result (the British won) and the experimenter then instructed them to
judge the likelihood of various outcomes as if they did not know what had
actually happened. Despite this instruction, when participants knew the
outcome, they exhibited hindsight bias. The bias was especially strong
when experimenters provided a deterministic cause of the British victory:
the superior discipline of their troops. But hindsight bias was practically
nonexistent when experimenters suggested a chance cause - a freak rainstorm. Likewise, hindsight bias was especially pronounced in a subset of
Northwestern football fans who, when asked after a game to recall their
earlier predictions, also generated causal explanations of the outcome, such
as "our defense shut them down" or "they missed a crucial field goal:' People feel most strongly that they always knew the results when they can construct a satisfying causal scenario that makes the outcome seem inevitable
in hindsight.

Hindsight bias is so powerful that it occurs even when people are ex
plicitly instructed to disregard the actual outcome of an event. It is as if
knowledge of the outcome becomes instantly integrated with other general
knowledge in semantic memory, and people simply cannot treat this new
bit of information any differently from other information relevant to the
judgment they are trying to make. That hindsight bias persists even when
people are explicitly attempting to ignore outcome knowledge has potentially important implications for everyday situations in which hindsight
bias occurs. When you seek a second medical opinion regarding a debatable diagnosis, you want the new doctor to take a fresh look at your condition, unbiased by the opinion of the first physician. But given the potent
influence of hindsight bias, knowing the first physician's opinion may inexorably influence the judgment of the second, even if the second doctor tries
to ignore what the first one said. This inevitable result is just what happened when doctors who received a diagnostic label for a particular case,
such as leukemia or Alzheimer's disease, together with instructions to ignore it, were asked to make an independent diagnosis. These physicians
were still more likely to make a diagnosis consistent with the label than
were others who made their diagnoses without the benefit of a label.

Something similar occurs among courtroom jurors. Suppose that the
prosecution introduces evidence from a seemingly incriminating telephone conversation, the defense objects to it, and the judge rules that the
evidence is inadmissible. He then sternly instructs the jurors to disregard
the evidence in their deliberations. Numerous studies have shown that
mock jurors placed in such a situation cannot disregard inadmissible evidence, even in the face of explicit instructions to ignore it: they are more
likely to convict than are jurors who never hear the inadmissible evidence.
The same holds for incriminating pretrial publicity that jurors are instructed to ignore. Once the evidence enters the memories of jurors, they
are biased to feel that they "knew all along" that the defendant was guilty.

Hindsight bias, then, is ubiquitous: people seem almost driven to reconstruct the past to fit what they know in the present. In light of the
known outcome, people can more easily retrieve incidents and examples
that confirm it. Recent evidence links this selective recall to the combined
influences of two forces: general knowledge that influences the perception
and comprehension of events, and vulnerability to misattribution.
Consider the following scenario. Barbara, a twenty-four-year-old single woman living in New England, meets an outgoing and intelligent man
named Jack in her graduate business class and begins working on a course
project with him. They begin to socialize after class, talking about school,
careers, and their mutual love of skiing. At one point they go to a restaurant, and Jack argues with a waiter and yells at Barbara, who then walks
home alone in tears. After the conclusion of the course, Jack and Barbara
stay out all night drinking and celebrating, and Barbara accepts Jack's invitation to spend a weekend at his parents' ski lodge in Vermont. The first
night, Barbara drinks wine at dinner and kisses Jack. After skiing the next
day, Jack takes Barbara out for a special dinner; they drink wine and Jack
holds Barbara's hand. After dinner, they return to the lodge, where Jack tells
Barbara she is sexy and that he loves her, and Barbara tells Jack that she
cares for him.

The psychologist Linda Carli asked Wellesley College undergraduates
to read a passage about Jack and Barbara, presented as a case history of a
woman who had been interviewed in a study of important life experiences.
Carli constructed two different endings to the story. After the part in which
Barbara told Jack that she cared for him, half of the students read that Jack
proposed marriage, whereas the other half read that Jack raped Barbara.
Two weeks later, all students were asked to rate the likelihood of alternative
endings to the story as if they did not know the ending, and also took a
memory test involving specific incidents that did or did not occur in the
story.

Carli found strong evidence for hindsight bias: students who read the
version ending in a marriage proposal judged the proposal as a more likely
outcome than students who read the version ending in rape, and vice versa.
Students who read the proposal ending tended to recognize falsely incidents that had not actually occurred, but are expected precursors of a marriage proposal, such as "Jack gave Barbara a ring:' "Barbara and Jack dined
by candlelight," or "Barbara wanted a family very much." But students who
read the rape ending tended to recognize, also falsely, possible precursors of
a rape, such as "Jack was unpopular with women," "Barbara was a tease:'
and "Jack and Barbara often went out drinking after work." Further, students' tendencies to misremember the precursors predicted the magnitude
of hindsight bias: more false memories resulted in more hindsight bias.

The results suggest that as students tried to reconstruct what had happened in the original passage, they activated general knowledge related to
the story ending that they had read - proposal or rape. Sometimes they misattributed this knowledge to the story, leading them to misremember
what happened and also to believe that they "knew all along" that the story
would end in a manner consistent with the one they read.

Hindsight biases are worrisome insofar as they can reduce or even
prevent learning from experience: if we feel that we knew all along what
would happen, then we may be less inclined to profit from the lessons a
particular event or incident can teach us. But at the same time, the comforting sense that we always knew the way things would turn out makes us
feel good about ourselves, inflating estimates of our own wisdom and prescience. This feature of hindsight bias no doubt contributes to its potency,
because self-enhancing biases are pervasive features of attempts to reconstruct the personal past.

I REMEMBER IT WELL.

In the 1958 musical Gigi, former lovers, played by Maurice Chevalier and
Hermione Gingold, reflect back over the years and recall their final date together. As illustrated in the song "I Remember It Well:' even though each
one remembers the occasion vividly, their recollections could hardly be
more different:

I can remember everything, as if it were yesterday.
HE: We met at nine.
SHE: We met at eight.
HE: I was on time.
SHE: No, you were late.
HE: Ah yes, I remember it well.

We dined with friends.
SHE: We dined alone.
HE: A tenor sang.
SHE: A baritone.
HE: Ah yes, I remember it well.

That dazzling April moon ...
SHE: There was none that night.
And the month was June.

The song continues with an ever-accumulating series of conflicting
memories. One of the pair must be wrong on each point, but neither one
ever backs off from his or her side of the story. Most couples can likely recall similar, if not so extreme, examples from their lives. At a recent party
during the December holiday season, a graduate student working in my
laboratory almost came to blows with her husband because of a memory
conflict over who made jelly doughnuts at last year's holiday get-together.
She recalled in vivid detail making the treats and serving them; so did he.

We are likely to give more credence to our own recollections of events
than to those of others when our memories readily spring to mind and are
accompanied by vivid, compelling details. We have direct access to these
qualities of our own recollections in a way that we never do for the memories of others, which can lead us to dig in and insist on the unique validity
of our own view of the world. This kind of egocentric bias contributes to
some of the disagreements that couples experience about their shared
pasts. Studies of married and dating couples have shown, for instance, that
each member of the couple tends to remember himself or herself as more
responsible for various kinds of incidents than the other. When asked to recall how much they contributed to deciding how money should be spent,
planning a vacation together, or similar activities, one spouse might claim
80 percent credit while the other claims 40 percent. Although both agree
that one of the pair was more responsible, one or both of them are claiming
too much credit for their own contributions. This egocentric bias occurs
even for negative incidents, such as shouldering too much responsibility
for causing arguments in the relationship. The bias likely occurs because
each member of the pair can more easily recall his or her own actions and
feelings than what the partner did or said. Laboratory studies have shown
that we tend to recall our own actions and words more readily than those
of others.

Egocentric biases in memory reflect the important role that "the self"
plays in organizing and regulating mental life. Many psychologists conceive
of the self as a richly interconnected knowledge structure - the sum total
of stored information about personal attributes and experiences. Numerous experiments have shown that when we encode new information by relating it to the self, subsequent memory for that information improves
compared to other types of encoding. If I ask you to think about whether
such attributes as "honest" or "intelligent" describe you or not, you are
more likely to remember those words than if I ask you to make the same
judgments about somebody else, such as a friend or a celebrity. Self-encoding also produces higher levels of subsequent memory than asking you to
elaborate on the words by focusing on their meaning or other properties
that are not directly related to the self.

But the self is hardly a neutral observer of the world. Individuals in
our society are motivated to think highly of themselves and often hold unrealistically flattering opinions of their abilities and achievements. Studies
summarized by the social psychologist Shelley Taylor and her associates indicate that people are commonly subject to "positive illusions" characterized by inflated estimates of self-worth. For instance, most people tend to
view desirable personality traits as more descriptive of themselves than of
the average person, but view undesirable personality traits as less descriptive of themselves than the average person. Because most people cannot be
better than average, for some of us this sunny self-assessment must be illusory. Likewise, people are more likely to attribute successes than failures to
themselves, and to attribute failures to forces outside the self.

The self's preeminent role in encoding and retrieval, combined with a
powerful tendency for people to view themselves positively, creates fertile
ground for memory biases that allow people to remember past experiences
in a self-enhancing light. Consider, for example, college students who were
led to believe that introversion is a desirable personality trait that predicts
academic success, and then searched their memories for incidents in which
they behaved in an introverted or extroverted manner. Compared with students who were led to believe that extroversion is a desirable trait, the introvert-success students more quickly generated memories in which they
behaved like introverts than like extroverts. The memory search was biased
by a desire to see the self positively, which led students to select past incidents containing the desired trait.

Similar processes operate in everyday situations in which people are
highly motivated to recount their pasts in ways that enhance current selfassessments. Do you recall what grades you obtained in high school
courses? Can you remember how many As and Ds appeared on your report
card? Chances are that you will recall more of the good grades than the bad
ones. When college students tried to remember high school grades, and
their memories were checked against actual transcripts, they were highly
accurate for grades of A (89 percent correct) and extremely inaccurate for
grades ofD (29 percent correct).

Divorce can also accentuate self-enhancing memory biases. Recently
divorced couples' retrospective assessments of their failed marriages reveal
that each member of the pair tends to portray the past from very different,
consistently self-serving perspectives. Looking back on why their marriage
ended, one man recalled that «all she wanted was money to put in the
bank;' whereas his ex-wife remembered that «My husband seemed to be
obsessed with making money:' Another man attributed his breakup to the
fact that he met another woman who was «younger and better-looking;'
whereas his ex-wife characterized the new woman as «a real bimbo;' recounting that «people were prone to using descriptions such as 'the elevator doesn't go quite to the top:"

Self-enhancing biases can also result from exaggerating the difficulty
of past experiences. Consider a situation in which you anxiously study for a
tough exam, take the test, and later find out that you passed. Just how anxious were you prior to the exam? Graduate students who recorded their
anxiety levels before taking an important set of comprehensive examinations were asked a month later to recall how anxious they were during the
pre-exam period. Students tended to exaggerate their pre-exam anxiety
levels; the memory bias was particularly pronounced in those who knew
that they had passed the exam. Recalling greater levels of anxiety than they
actually experienced enhanced the students' sense of accomplishment, increasing pride and confidence in their abilities to cope with adverse events.
Blood donors show a similar memory bias, retrospectively inflating their
levels of predonation anxiety in a way that heightens their sense of bravery
in overcoming obstacles to accomplish a courageous deed.

People sometimes deprecate past selves to maintain and enhance a favorable view of the current self. «Of all the lives that 1 have lived, 1 would
have to say that this one is my favorite;' the actress Mary Tyler Moore reflected in a 1997 magazine interview. «I am proud that 1 have developed
into a kinder person than 1 ever thought 1 would be. 1 am less critical than 1
ever was and, as a result, I'm less critical of myself:' Perhaps Tyler Moore
has indeed changed for the better over time. But by recalling her past self as
less kind and more critical than she is now, suggests the psychologist Michael Ross, she enhances the value of her present self. Ross finds that people
generally speak more favorably of present selves than of past selves. As with
Tyler Moore, this inclination could reflect either a genuine improvement
over time or a tendency to deprecate past selves. Consistent with the latter possibility, a substantial majority of college students and middle-aged
adults rate their present selves, but not past selves, as above average relative
to their peers. As noted earlier, a large majority of people cannot be above
average compared with their peer group. These results therefore suggest
that people inflate estimates of current self-worth by deprecating the way
they were in the past.

Egocentric memory biases, then, are reflected in several related maneuvers - selective recall, exaggerating past difficulties, and deprecating
past selves - that surround the present self in a comforting glow of positive illusions.

WHISTLING VIVALDI.

When the African American journalist Brent Staples arrived as a student at
the University of Chicago, he enjoyed walking near the lakeshore at night.
Staples became unnerved one evening when he noticed that a white businesswoman, suddenly aware of his presence on the street, walked away
quickly and then began to run. "I'd been a fool;' reflected Staples. ''I'd been
walking the streets grinning good evening at people who were frightened to
death of me:' Attempting to ease concern that he was stalking white pedestrians or was otherwise ill intentioned, Staples started whistling Vivaldi's
The Four Seasons to signal that he was a benign stroller. "The tension
drained from people's bodies when they heard me;' Staples recalled. ''A few
even smiled as they passed me in the dark."

Staples whistled Vivaldi because his presence activated in others'
memories a powerful stereotype that biased white strangers' perceptions
of him: when walking on a quiet street at night, a black man poses danger.
The resourceful Staples came up with an effective method to avoid being
viewed in such stereotypical - and erroneous - terms.

Stereotypes are generic descriptions of past experiences that we use to
categorize people and objects. Many social psychologists think of stereotypes as "energy-saving" devices that simplify the task of comprehending
our social worlds. Because it may require considerable cognitive effort to
size up every new person we meet as a unique individual, we often find it
easier to fall back on stereotypical generalizations that accumulate from
various sources, including discussions with other people, printed and electronic media, and firsthand experience. Though relying on such stereotypes may make our cognitive lives more manageable, it can also lead to
undesirable outcomes: when a stereotype diverges from reality in a specific
instance - as happened with Brent Staples - the resulting biases can produce inaccurate judgments and unwarranted behavior.

The great social psychologist Gordon Allport was one of the first psychologists to recognize how the dual nature of stereotypes contributes to
racial biases. While acknowledging that stereotypes help us to categorize
the world, Allport held that "we often make mistakes in fitting events to
categories and thus get ourselves in trouble:' In his classic 1954 book, The
Nature of Prejudice, Allport foresaw quite dearly the situation that Brent
Staples would confront decades later. "A person with dark brown skin will
activate whatever concept of Negro is dominant in our mind," contended
Allport. "If the dominant category is one composed of negative attitudes
and beliefs we will automatically avoid him, or adopt whichever habit of rejection is most available to us."

Allport's assessment was especially prescient because recent research
has underscored that stereotypical biases can occur automatically, outside
of conscious awareness. Early evidence for this view came from experiments that activated stereotypes by presenting words too quickly to register
in conscious perception (a procedure known as subliminal priming). After subliminal priming with words intended to activate a stereotype of
"blacks;' such as welfare, busing, and ghetto, white American students were
more likely to judge an imaginary male of an unspecified race as a hostile
person than when they were primed with neutral words. Further, the biasing effect was just as powerful in those students who expressed little racial
prejudice on a questionnaire as in those who overtly expressed considerable racial prejudice.

This latter finding is particularly troubling, because it suggests that
even people who consciously experience little prejudice automatically activate stereotypical biases. But more recent results from a British study point
toward differences between high- and low-prejudice individuals. Like
American students, high- and low-prejudice white British students were
both biased to see a race-unspecified person as hostile after subliminal
priming with negative words that directly activate a racial stereotype,
such as drugs, nigger, rude, and crime. But only high-prejudice individuals
showed a biasing effect after subliminal priming with neutral words that
activate the general category of "black people;' such as blacks, colored, afro,
and West Indians.

Stereotypical biases can also result in disturbing tendencies for people to "remember" hearing about nonexistent black criminals. Mahzarin
Banaji and her coworkers at Yale University showed college students male
names, indicating that some might seem familiar because they were names
of criminals who appeared recently in the media. Although none were actually names of criminals, the students were almost twice as likely to identify stereotypically black names (Tyrone Washington, Darnell Jones) as
those of criminals compared to stereo typically white names (Adam McCarthy, Frank Smith). The bias occurred even when people were instructed
that "people who are racist identify more black names than white names;
please do not use the race of the name in making your judgment."

Bias effects are not restricted to racial stereotypes. In another series of
studies, Banaji and collaborators exposed people to names of famous and
nonfamous people, and later asked them to judge whether these and other
names were famous or not. Previous studies had shown that after seeing
nonfamous names, people sometimes later mistakenly classify them as famous. The nonfamous names seem familiar because they were presented
earlier in the experiment, but participants forgot where they encountered
the name a misattribution error similar to those we considered in Chapter 4. In Banaji's experiment, people were far more likely to make this "false
fame" error for male names than for female names. A gender stereotype men are more likely to be famous than women - biased participants to
make erroneous claims about the supposed fame of made-up male names.

A case can be made that such stereotypical biases are defensible and
even reasonable. After all, in our society men are more likely to be famous
than women; likewise, a higher proportion of black than white men are in
prison. The latter consideration probably motivated the behavior of the
night strollers who conspicuously avoided Brent Staples: the University of
Chicago area where he lived borders on largely black and crime-ridden
neighborhoods. Considered in statistical terms that apply to groups of people - men and women, blacks and whites - stereotypical biases are not
necessarily erroneous. The problem arises because people are sometimes
willing to act on these biases in cases in which they are entirely unwarranted, resulting in what Banaji calls "guilt by association" rather than
"guilt by behavior": individuals are perceived negatively based on their
membership in a group rather than because of their specific behaviors or
attributes.

Activated stereotypes bias not only how we think and behave; they can
also influence what we remember. If I tell you that Julian, an artist, is creative, temperamental, generous, and fearless, you are more likely to recall
the first two attributes, which fit the stereotype of an artist, than the latter
two attributes, which do not. If I tell you that he is a skinhead, and list some
of his characteristics, you're more likely to remember that he is rebellious
and aggressive than that he is lucky and modest. This congruity bias is especially likely to occur when people hold strong stereotypes about a particular group. A person with strong racial prejudices, for example, would be
more likely to remember stereotypical features of an African American's
behavior than a less prejudiced person, and less likely to remember behaviors that don't fit the stereotype. This tendency can create a self-perpetuating cycle in which a stereotype biases recall of congruent incidents, which
in turn strengthens the stereotypical bias.

Stereotype bias also tends to occur when we don't make an effort to
consider an individual's particular characteristics because we are mentally
preoccupied with other matters. In controlled experiments, for instance,
stereotype bias is most pronounced when people are given difficult tasks to
carry out at the same time that they form impressions of people. You are
most likely to recall only that Julian the artist is creative and temperamental
if, when you first meet him, you are devoting most of your attention to
thinking about an important meeting, or an exam you will soon be taking.
When you can devote more cognitive effort to sizing up Julian as an individual, you may actually recall more information that is incongruent with
the stereotype. If you notice that Julian seems unusually even-keeled, for
instance, you may wonder why he is so different from your stereotypical expectation of a temperamental artist. As a result of carrying out elaborative
encoding to resolve the apparent discrepancy, you later remember clearly
Julian's even-keeled demeanor.

When events unfold in a way that contradicts our expectations based
on stereotypes and related knowledge of the world, we may be biased to
fabricate incidents that never happened in order to bring our memories in
line with our expectations. Consider two versions of a story about a man,
Bob, who dearly wanted to marry his girlfriend, Margie, but did not want
children and was anxious about how Margie would react if he told her. In
one version of the story, Margie was thrilled to hear that Bob wished to remain childless because this desire fit well with her career plans; in another
version, Margie was horrified because she desperately wanted children.
Now consider two possible endings: Bob and Margie married, or Bob and
Margie ended their relationship.

If you had read that Margie was thrilled by Bob's disclosure, then
based on your general knowledge of relationships, you would expect them
to marry and find it surprising if they split up. But if you had read that
Margie was horrified by Bob's disclosure, you might expect a breakup and
be surprised if they married. Experiments have shown that when trying to
recall the story, people who received incongruous endings mistakenly remembered critical incidents in a way that made sense of the outcomes. For
instance, participants who read that Margie was horrified, and then learned
that Bob and Margie married, recalled incorrectly, "They separated but realized after discussing the matter that their love mattered more." But participants who read that Margie was thrilled, and later learned that the couple split, incorrectly recalled such incidents as, "There was a hassle with one
or the other's parents," or "They disagreed about having children:'

Much as the Ministry of Truth in Orwell's 1984 revised the historical
past to fit its current precepts, general knowledge biased story recall so that
memory fit neatly with expectations. In 1984, responsibility for revision
and fabrication fell on the shoulders of workers within the Ministry of
Truth, such as the novel's protagonist, Winston Smith. In the world of
memory, revisionist biases have been linked with one of the most puzzling
subsystems in the human brain.

THE BASIS OF BIAS.

In the late 1960s, neuropsychologists described an arresting syndrome
that immediately captured the imagination of scientists and the general
public alike. Patients who had undergone surgical separation of the left and
right cerebral hemispheres as a treatment for intractable epilepsy - "splitbrain" patients - behaved as if they housed two minds in a single body.
The left hemisphere handled language and symbols, the right specialized in
nonverbal information such as images and spatial locations. Though these
patients seemed normal in casual conversation and social interaction, careful psychological testing revealed that situations could be devised in which
each hemisphere digested incoming information without awareness of
what the other was experiencing.

Despite its ignorance of happenings in the right hemisphere, the left
hemisphere is nonetheless quite adept at coming up with various explanations and rationalizations for the strange situations that result from its surgical disconnection. The Dartmouth neuroscientist Michael Gazzaniga,
who pioneered much of the research on split-brain patients, used clever ex158 peri mental procedures to create conflicts between the left and right hemispheres to reveal the left brain's propensity for explanation and rationalization. For instance, after flashing the command "walk" to a split-brain
patient's right hemisphere, without the knowledge of the left hemisphere,
the patient would get up as instructed. When asked why he was walking,
the patient - now relying on the verbal left brain - rationalized that he
was going to get a soda. In another classic demonstration, Gazzaniga exposed a picture of a snow-covered house to the right brain, and a chicken
claw to the left. The patient was instructed to pick (from several choices) a
line drawing of an object that related to the picture he saw. The patient's
right hand (controlled by the left hemisphere) chose a rooster to match the
chicken claw, whereas his left hand (controlled by the right hemisphere)
chose a snow shovel to match the winter scene. The patient, faced with the
bizarre sight of his two hands pointing to different drawings, consulted his
verbal left hemisphere (which had no knowledge of the winter scene presented to the right brain) and immediately offered an explanation. He
chose the shovel with his left hand, the patient claimed, because it could be
used to clean out the chicken coop! Innocent that the left hand actually
chose the shovel because it fit the winter scene shown to the nonverbal
right hemisphere, the left brain confidently - but erroneously - generated an after-the-fact rationalization that made sense of the otherwise bewildering choice.

Based on these and other similar observations, Gazzaniga postulates
that the left brain contains an "interpreter" that is continually drawing on
general knowledge and past experience to try to bring order to our psychological worlds. These activities can produce memory biases not unlike
those considered earlier in this chapter. For instance, Gazzaniga and his
colleague Elizabeth Phelps showed split-brain patients slide sequences of
such everyday activities as a man getting up for a day's work. Later, they
tested the memories of the left and right hemispheres for incidents shown
earlier, such as the man looking at his alarm clock, and novel incidents that
had nothing to do with the studied sequences, such as the same man fixing
a television. Most important, each hemisphere was also asked about incidents that fit the stereotype (or "schema") of getting up for work, but that
did not actually appear in the initial slide sequence - sitting up in bed,
brushing teeth, and the like.

The left hemisphere often falsely recognized novel incidents that were
consistent with the stereotype, whereas the right hemisphere hardly ever
did. The left brain interpreter was at work again, showing a bias to respond
on the basis of general knowledge about activities usually involved in getting up for work. Though the left hemisphere's responses made sense in
general terms - people do typically sit up in bed or brush their teeth when
they get up for work - they were wrong when applied to the particular
slides shown on this specific occasion.

The resemblance to the stereotypical biases considered earlier is striking. The left brain interpreter relies on inferences, rationalizations, and
generalizations as it tries to relate past and present, and in so doing probably also contributes to consistency, change, hindsight, and egocentric biases. The interpreter may help to confer a sense of order in our lives, allowing us to reconcile our present attitudes with our past actions and feelings,
generating a comforting sense that we always knew how things would turn
out, or enhancing our opinions of ourselves. But it also has the potential to
lead us down the path of delusion. If the facile explanations and rationalizations offered by the interpreter generate powerful biases that prevent us
from seeing ourselves in a realistic light, we are clearly at risk for repeating
past failures in the future.

Fortunately, the left brain interpreter is balanced by systems in the
right hemisphere that are more attuned to the constraints of the external
world. In Phelps and Gazzaniga's memory study, for instance, the right
brain claimed to remember only the exact incidents it witnessed, and almost never falsely recognized similar events that hadn't happened. In fMRI
studies led by Wilma Koutstaal in my laboratory, we discovered that part of
the right visual cortex is sensitive to whether an identical object is presented on two occasions (the same picture of a table), compared with two
different examples of the same object (pictures of different tables). But the
left visual cortex responds similarly whether objects shown on two occasions are identical or merely alike.

The right hemisphere's proclivity for responding on a literal basis can
help to keep in check its more expansive and error-prone cerebral neighbor. In Orwell's 1984, the Ministry of Truth enjoyed sovereign rule, unfettered by any countervailing forces; the result was a totalitarian disaster. The
left-brain interpreter, left to its own devices, might well produce a similarly
calamitous outcome in individual minds: unchecked bias and rationalization could lead us to a bottomless abyss of self-delusion. Happily for our
species, however, the brain has engineered a system of checks and balances
missing in Orwell's nightmare vision. Still, the various forms of bias are so
deeply embedded in human cognition that few good remedies exist for
overcoming or avoiding them altogether. Perhaps the best we can do is to
appreciate that current knowledge, beliefs, and feelings can influence our
recollection the past, and shape our impressions of people and objects
in the present. By exercising due vigilance, and recognizing the possible
sources of our convictions about both past and present, we can reduce the
distortions that arise when memory functions as a pawn in the service of its masters.