Chapter 4 : Schema Theories
The Nature of Schemata
We categorize people and then use those categories to draw inferences about them. Sometimes this is a
passive and m indless sort of thing, but oth er times it is active and thoug htful. In addition to ca tegory-based
knowledge, we usually have more immediately observed information about people’s behaviors, traits, and
appearance, and so we also have to integrate all this pre-existing and incoming information. Since the time of the
ancient Greeks, most philosophers and psychologists have agreed that we often process information in a “topdow n” fas hion, th at wh at we “ know ” dicta tes wh at we “ see” eve ry bit as m uch a s the rev erse. Wh at we c all
experience is alw ays a fun ction of bo th prior theories held in mem ory and present inpu t from ou r sensory
appara tus. Th ere are no naked experienc es. Tho se that involv e other p eople are es pecially like ly to be co gnitively
adorn ed just b ecaus e behav ior mu st be inter preted in terms o f contex tual fac tors an d inferred intentio ns.
In modern cognitive psychology, the notion of schema has been used to represent the prior knowledge
part of the equation. Schemas may loosely be defined as theories we have about categories, and they function as
frameworks for understanding what we see and hear. Some of these theories may be quite impoverished – I do
not, for example, have any sort of elaborate theory about earthworms – but for important categories, say
occupations or gender, I have extensive knowledge that is bound together by quite complex theoretical ideas.
When a student observes a professor giving a lecture, she will have schemata for the professor role, for the
lecturing process, and for various other categories (male, Hispanic, young) that this particular professor fits.
Thus, stereotypes can be thought o f as schemata
The Importance of Prior Knowledge
It is not obvious that we need an elaborate psychological construct, such as schema, to help us
und erstan d how people co nstru e their wo rlds. For examp le, my ca t whic h screa ms at t he bac k doo r to be let in
seems to have an exceptional fear of being hit by the screen door as it opens. And so when she hears someone
approaching the door, she retreats dow n a couple of steps m akes exactly one full turn a nd waits for the screen
door to open before she darts in. Now it is quite possible that this animal has a well developed theory about the
indignities of being hit by doors and the necessity to ward off such evils with retreats and turns. But it seems
more likely given other evidence tha t she has no more than a couple dozen functioning co rtical cells, that her
behavior is entirely a matter of conditioning and ma y not be represented cognitively at all. At least some, perhaps
muc h, of hu man thou ght an d beha vior m ay also be due to con ditionin g and other fo rms of p assive learning . I
need h ave no theory abou t anyth ing to b e afraid o f the large man in the d ark alley way.
Dictionaries vs. Encyclopedias
Still, when all is said and do ne, it has seemed reason able to many m odern psycho logists that we need
some term to re present o ur pas t know ledge ab out pa rticular c oncep ts and cons tructs. S ome m odels c onstru e this
past know ledge as bits of information attached to co gnitive nodes, and such asso ciation models in w hich features
are linked to categories have been popular, especially in the stereotype area (Stangor & Lange, 1994); our
definition of stereotypes (Chapter 1) incorporates this assumption. Certainly we do have such knowledge
representation. So just I might report that dogs have tails, bark, and fur and that cars have transmissions, engines,
doors, and seats, I could believe that men are assertive, logical, and emotionally impoverished. This might be
called the dictionary view of knowledge -- our knowledge of a thing consists of several features that are more or
less true of the object in question. The dictionary view does not require that we have sophisticated knowledge
about the concepts under question; dictionary entries may be passively acquired, mechanically stored, and
unreflectively used. If stereotypes are nothing more than relationships between traits an d groups, then w e need
not assume much about the theories of the people holding them.
However, everyday observation suggests this is a crude and limiting view because often the theoretical
basis of our knowledge seems crucial. For example, almost any adult in our society can discuss the relationships
between the phy sical feat ures of a car; we have s ome s ense of w hy do ors are w here ther e are, wh ere the en gine is
and wh y, and the wa ys the transmission affects speed. All of this affects my driving beha vior and reaction s to
problems in fundamental ways. In other words, knowledge about a car is more like an encyclo pedia entry, with
Chapter 4: Schemas / Version of November 10, 2002
1
notions abo ut how th e various parts of the kn owledge base a re interrelated and how each, in turn, is related to
other things through a series of what we might call hyperlinks. Accessing information about engines and how
they work might put me in the frame of mind to turn to a related "entry" on power or gasoline or the importance
of oxygen in combustion. In the last chapter I argued that many, probably most concepts, and certainly most
concepts that are foundations for stereotypes, enfold a great deal of theoretical knowledge. We know what
features are impo rtant fo r variou s categ ories, wh y they a re impor tant, a nd ho w they are related . Of cou rse, it
remains to be seen wh ether we ga in any levera ge on ou r understa nding o f stereotypes by assum ing that th ey are
theories of sorts abou t group s of peop le.
This is not an either-or situation. It has long been recognized (e.g., Sloman, 1996) that there are at least
two forms o f reasoning. On e is based on ass ociations and is usually quick an d fairly automa tic. The other is rulebased or theo ry-bas ed an d takes some cogn itive effort. T here is ev idence t hat bo th form s of reas oning occu r in
the us e of stereo types (e.g., S mith & DeCo ster, 200 0). For ex ample, w hile som e inferenc es surely take tim e (e.g.,
inferring someone’s intelligence from the way she is dressed), others seem fairly automatic at least in the sense of
not being conscious (e.g., inferring that the person standing at the head of a classroom is a professor). While the
focus of this chapter will be on theory-based processing, we need to remember that some forms of stereotype
inference do not even require categorization, let alone high level inferences. For example, prototypic AfricanAmeric an fac es cue n egative e valua tions (L ivingst on & B rewer, 20 02) an d stereo types (B lair, Judd , Sadler, &
Jenkins, 2002) independently of whether the person categorizes the face as African-American. Also
affective/emotional reaction s to features associated with salient categories ma y kick in quickly, well before even
implicit catego rization h as likely taken place (Stapel, K oomen , & Ruy s, 2002).
Functions of Schemata
Our theories or knowledge structures -- our schemata -- aid in the recognition, interpretation, and
labeling of stimu li, affect mem ory for in forma tion, pro vide de fault v alues fo r missin g inform ation, a nd gen erally
lead to efficient processing of schem a related material (Fiske and Ta ylor, 1991). People with w ell developed
theories or sch emata abou t some dom ain are, in effect, experts abou t that d oma in. My k now ledge ab out ca rs is
extremely limited (although not as limited as that of small children). I can identity some major parts of the
engine, recognize the difference between tires and exhaust systems. Indeed, I could probably write pages and
pages of information about what I know about cars even though I lack, and lack by a large amount, the
know ledge to d iscou rse intelligen tly abo ut the in ner wo rkings o f engines or to repa ir them. M y mech anic, w ho is
expert abou t such m atters, has a much more dev eloped theo ry abou t all this and could certa inly offer a mo re
intelligent (and probably concise) statement about the workings of cars.
Tradition ally psych ologists ha ve assu med tha t all this cogn itive work o f attention, lab eling, memo ry
reduction, and inference-drawing is necessary to reduce the amount of information that is available about our
world, and especially the world of other people. There is no doubt that schemata do help us process information
and in that sen se make our interac tions with the w orld more efficient. So, for example, activating a sch ema wh en
forming impressions of people uses fewer cognitive resources than forming impressions without such weapons
(Macrae, Milne, & Bo denhau sen, 1994), and generally having an available schema or stereotype facilitates
processing stereotype releva nt inform ation qu ickly (Kaw akam i, Young , & Dov idio, 2002 ).
However, schemata are also important because they enrich our understanding of the world (Oakes &
Turner, 1990) in that they allow us to infer features that we cannot immediately perceive. Behavior that might
seem incomp rehensible “in the raw” may ma ke perfect sense in the context of a relevant sc hema. For exam ple,
we might not understand the behavior of a man who rushes out of an important business meeting until we know
that he has just received a call that one of his children h as been injured at sch ool. Perhaps even m ore to the point,
our schemata allow us to go beyond the information given (Bruner, 1957). Once I know that a computer is using
Windo ws98 as an o perating sys tem, I can predict how the com puter will respon d to a w ide range o f keyboard
comm ands and m ouse c licks even thou gh I m ay no t fully un dersta nd th e progra m tha t is runn ing.
It is true th at ma ny pred ictions from sc hema mod els can be ma de mo re easily fro m less a ggressiv ely
theory-driv en acco unts (e.g., Alba & Has her, 1983 ) and tha t schema models a re often bloate d, vagu e and “u nderspecified” (Carlston & Smith, 1996). However, such theories historically have suggested hypotheses and lines of
research not easily derived from other models, and even though schema models are not parsimonious they do
account for a wide range of phenomena. In particular they provide safe haven for the kinds of theoretical
reason ing tha t I claim often (bu t not in evitably ) accom panies stereoty pes.
Chapter 4: Schemas / Version of November 10, 2002
2
Schemata and Information Processing
Stages in Processing Information
The schema notion has generally been paired with the additional idea that we process information about
others in a series of somewhat discrete but overlapping stages. We will consider four: attention, labeling and
interpretation, memory, and inferences.
Suppose, at a crowded party, I introduce you to a person named Scott, a man I've been telling you has a
lot in common with you, and a guy I think you'll like. I now move away to "let you two get acquainted." From
your p oint of v iew you 're stuck with S cott bec ause y ou do n't kno w an yone els e at this p arty, an d so yo u look him
over phy sically and metaph orically. Wh at do yo u pay a ttention to? Surely a lot d epends o n the sch emata y ou try
out. If y ou're th ink of S cott as a poten tial rom antic p artner, y ou m ight imm ediately n otice tha t he is tall, fa irly
attractive, and has a nice baritone voice. If you are the type of person who prefers intellectuals to more social
types, you may particularly notice that his tie is unfashionable, his shoes are unshined, and that he is wearing
thick glasses, cues for intellectual interests in you r schemata o f the intellectual. As you con tinue to talk with Sco tt,
you may ask him some questions or try to steer the conversation in a revealing direction. And the categories you
employ may change. At first, you m ay have ca tegorized Scott as a b it of an intellectual because of his ap pearance,
but then you find out that he is a lawyer, a category of people you do not associate with intellectualism or much
like. Why did I introduce you to a lawyer? And how can a lawyer dress as he does? But then you discover that he
works for a local poverty agency, helping abused women with legal problems, and you begin to think of him as a
"liberal", or a "com passion ate guy". So mewh ere along th e line, Scott ma y reveal tha t he lives in a fas hionab le part
of town, and you now have to figure out h ow a low paid lawyer can live there. You infer that he has in herited
money, and you begin to test whether he fits your schema of a "guilty liberal". At times during the conversation
you may wonder whether he is, gay, romantically involved with someone, a sports fan, or a lover of classical
music. The point is that you will pay attention to different aspects of his behavior depending on which of your
schem ata ho lds cen ter stage at any given m omen t.
So we wo uld expect attention to be d irected by our prior know ledge. Ultimately as your attention , driven
by va rious s chem ata, yield s kno wledg e abou t Scott, y ou w ill label him , will place h im in on e, proba bly seve ral,
catego ries. Scot t, you d ecide, is a " yuppie t ype" des pite his job , or a "H arvar d type", o r even a "sensitiv e male
type". Your interpretations of his beh avior wou ld also be affected by these decision s. His extended com ments
about the present administration in Washington may be labeled quite differently depending on whether you
think he is a guilty liberal or is trying to impress you. His sloppy dress will mean something different if you find
that he has just co me from wo rk where he mu st "dress down ", or is simply indifferent to his appearance.
Your mem ory for Scott and his behavior w ill surely be affected by how you label him, by wha t schemata
you use to u nderstand h is behavior. it would b e easy to imagine tha t when, a few m onths later, you try to
remember Scott, you would rem ember some featu res but not others. If you have decided th at he is a guilty
liberal, you may recall his political views but not his comments about his musical tastes or how nicely he smiled.
On the oth er hand, if you w ere thinking of him as a friend or romantic partn er, you might well recall his smiles
and d ress bette r than his intric ate politic al theo ries.
Finally, we wou ld expect that your kn owledge abo ut the type of person Sco tt is will affect what you infer
about him . You migh t easily imagine that he w as not in a fraternity in college, but tha t he was active in liberal to
radical political organizations. You suspect he majored in political science or sociology, and are surprised that he
was a physics major. Why would a physics major go to law school and work for a poverty organization?
Obviou sly although w e have been discu ssing schema ta, most of wh at we have b een saying abo ut Scott
really applies to stereotypes. In your attempts to get a handle on Scott, you are clearly drawing on your
stereotypes about lawyers, people who live in certain areas, liberals, and folks who dress a particular way.
Obviou sly while your stereotypes allow you to infer all mann er of things abou t Scott you nev er observe, they
also lead to errors. Some a re merely amusing su ch as you r inference that Scott was a sociology m ajor in college,
but other errors can be embarrassing and sometimes harmful. If you assume that Scott must be neurotic because
he is a g uilty libera l, you m ay reject h im befo re you h ave a c hanc e to disc over th at you could be goo d friend s.
Let us now review the major functions of schemata in information processing.
Attention
Attention and Identification: Schemata h ave major effects on atten tional processes, althou gh the effects
are com plex. Wh en you enter a ro om fo r the first tim e, you p robab ly direct yo ur atten tion to inform ation t hat w ill
allow you to discover wha t kind of a room it is; you would focu s on informa tion that wou ld allow you to
Chapter 4: Schemas / Version of November 10, 2002
3
discriminate bedrooms from living rooms. In this case it would not do you much good to glance at the ceiling or
the floor, bedrooms a nd living room s usually ha ving abou t the same requirements in those areas. You might,
instead, examine the room for evidence of mirrors, certain types of furniture, and the like. Sometimes,
information that is especially relevant to our interests, goals, or active schema seems to leap out at us
perceptually. In one experimental demonstration, Macrae, Stangor, and Milne (1994) primed stereotypes by
havin g sub jects list traits relevan t to the s tereotyp e. Then the su bjects w ere asked to recog nize perc eptually
degraded w ords that were related or u nrelated to the stereotype, and th ose who h ad the stereotypes primed
foun d it easier to recogn ize the st ereotype relevan t inform ation.
Attention to Inconsistencies : Once you have securely identified that which is before you cognitively,
you can relax attention from these high ly diagnostic and critical features and afford to pay more attention to
whatever features seem strange or inconsistent. For example, I would imagine that finding a commode in the
middle of wha t otherwise seems to be a normal living room might attract and hold your a ttention for some time;
you might even be inclined to ask your host a guarded question or two about the appliance. You might find
yours elf sneak ing glan ces at th e blunt and u gly fingers of a su rgeon o r the mis shap en teeth o f a dent ist. Simila rly
when you m et our o ld friend, S cott, yo u paid particu lar atten tion to (or sou ght ou t) inform ation t hat w ould h elp
you res olve w hat yo u tho ught were inc onsis tencies b etween his profe ssion (lawyer ) and a ttire (scru ffy).
In a study on attention to social features White and Carlston (1983) asked subjects to watch a videotape of
two people who sat at opposite ends of a library table engaged in separate conversations. The audio part of the
tape was rigged in such a way that subjects heard one of the conversations more loudly and clearly than the
other, b ut som e subjec ts cou ld shift th e focus of atten tion from one co nversa tion to the oth er by ch angin g their
relative lo udn ess. Su bjects w ere given trait infor matio n abo ut on e of the a ctors (k ind or h onest), a nd initia lly
subjects spent mo re of their time monitoring the beha vior of the person for wh om they ha d a personality
expectation, presumably to confirm their expectations. During the first several minutes this target person
performed behaviors that neither confirmed nor disconfirmed the expected trait. Consequently, over time the
preference for monitoring the person for whom they had an expectation gradually decreased, and subjects spent
relatively more of their time listening to the person for whom they had no prior trait information. However, at
one point, the target person began to perform a behavior that was inconsistent with the kind or honest
expectation, and at that point attention to that person markedly increased. What seems to have happened is that
subjects selectively monitored the perso n for whom they had an expectation to see if his behavior co nfirmed it.
When th e behavior turned out to be relatively uninform ative, they "lost interest" and began to shift attention to
other matters. However, when the person began to perform in unexpected ways, they once again gave much
more o f their atten tion to this pers on.
Several other studies also show that we especially attend to information inconsistent with our
expecta tions a nd sc hema ta (e.g., Ba rgh & T hein, 19 85; Fisk e & Neu berg, 19 90; Fried man , 1979 ; Hilton , Klein, &
von Hippel, 1991; Stern, Marrs, Millar, & Cole, 1984). The implication is that when people have a stereotype
about a gro up, they will be inclined to mo nitor the behavior of gro up memb ers to see whether their behav ior fits
the stereotype. If it do es, then atten tion ma y be given to other ma tters. If the process stops here (a s it often does ),
the perceiver will only know that the target belongs to a category or two and perhaps that she has some relevant
traits. However, if the perceiver is motivated to process information about the target more carefully, he may be
struck by stereo type inco nsisten t inform ation t hat is like ly to cap ture an und ue am oun t of atten tion. Th is
sugg ests tha t people m ay be pa rticularly apt to pic k up su perficial w ays in w hich o thers be have as their
stereotypes say they should. In cases where interaction is limited and basically stops after such encounters,
superficial behaviors may lead to stereotype confirmation. On the other hand, when the person does perform a
behavior inconsistent with the stereotype, additional attention and presumably processing will be given to that
behav ior. Ho w this incon sistent in forma tion is h andled is an im portan t theme for the res t of this c hapter.
What Captures Attention? Up to this point we have been arguing that attention to detail is controlled by
one's schemata. At different points in our information processing schema-consistent or schema-inconsistent
information will receive priority in processing. However, attention may also be affected by certain features of
stimuli themselves as well as our own motives.
Contrast: One o f the mo st obv ious fa ctors th at affects our at tention is the inten sity of stim uli or th eir
difference from the backg round . Regardless of wha t schema ta are active fo r you at a ny given mom ent you are
likely to pay more attention to the woman in the bright green dress than to her companion who is dressed in a
muted brown. You will probably steal an occasional glance at the man in the corner who is waving his hands
wildly or who is talking quite loudly. It is usually stimuli that stand out from their surroundings that capture the
Chapter 4: Schemas / Version of November 10, 2002
4
most attention (McArthur & Ginsberg, 1981). The woman's bright green dress would surely attract less attention
if the oth er wom en in th e room were als o wea ring brigh t and liv ely color s. The m an w aving his arm s wou ld
warrant no more than an initial glance in a room of heavy gesticulators. Being different is a magnet for attention
from others. For example, as we saw in the last chapter, Kanter (1977a) argued that people in a solo status (she
focused on females in male dominated positions) feel as if they are under constant scrutiny and the research
evidence does support the idea that tokens often get more than their share of attention
Another kind of novelty rests on departures from general norms. White Americans who travel to areas of
the wo rld wh ere white skin an d relativ e tallness are quite unco mmo n, often report th at nat ives, espe cially
children, scrutinize them carefully.1 Those wh o are physically different (e.g., disabled) often complain th at other
people stare at them.2 A study by Langer, Fiske, Taylor, and Chanowitz (1976) confirms that when subjects’
thought th eir staring could not be d etected, they did look mo re at a picture of a wom an in a leg brace; females
(but not males) stared more at a picture of a pregnant woman. While such attentiveness is often perceived to be
rude, it m ay often be mo tivated by no thing m ore tha n curio sity abo ut a rela tively un familiar s timulu s.
Situationa l Circumstan ce: In everyday life, surely one of the m ore important reason s we pay attention to
some things and not others is their relative availability to our sensory apparatus. We usually pay more attention
to people who are physically close rather than distant because it is easier to monitor them. People sitting across
from us are mo re salient than those sitting to o ur sides (Fiske, Kenny, & T aylor, 1982; Tay lor, Crocker, Fiske,
Sprinzen , & Wink ler, 1979; Ta ylor & Fiske, 1 975). 3
Motive s and G oals : It is an o bviou s but im portan t point th at often our at tention is cont rolled by our go als
and motives (Erber & Fiske, 1984; Neuberg & Fiske, 1987; Ruscher & Fiske, 1990). A student visiting his professor
to plead for additiona l points on an exa m surely pays clos e attention to how much th e professor smiles or grunts
as he rereads the exam, and he will pay special attention to what the professor says in hopes of finding an
opening to advance his case further. It is the curse of people who are waiting for decisions about promotions
within an organization, that every smile, frown, and minute bit of bodily behavior of superiors is captured and
interpreted for any sign of how the decision will go.
Consequences of Attention: The effects of attention are often obv ious. Unless on e pays attention to
some stimulus, information about it is not available for further processing. You will never form an impression of
the wom an acro ss the roo m, let alone th ink abou t her in stereotyp ic terms, un less you first o bserve her beh avior.
Attention and Attributional Processing: Howev er, there are effects of differential attention even when
enough attention is paid to several people to remember them. Imagine that you met two people at a party, and
that for some reason you paid far more attention to Hank than to Sara. How might we expect that extra attention
to affect your perceptions of them ? One specu lation is that people to wh om we give m ore attention are often seen
as cau sally more a ctive and powerfu l in a given situ ation. As we hav e know n for som e time (e.g., Schneid er,
Hast orf, & Ells worth , 1979 ), most o f us ten d to see o thers as auth ors of th eir own behav ior. We s ee people
responsible for their behavior in the sense that we assume that their intentions, attitudes, motives, and abilities y
cause their behavior, and several studies find that people to whom we attend are seen as particularly prominent
or casually active (Fiske, Kenny, & Taylor, 1982; McArthur & Post, 1977; Taylor, et al., 1979). Others find that
visual sa lience leads to m ore extreme trait ra tings (Tay lor, et al., 1979; M cArthu r & Solom on, 197 8).
Attention and Inhibition: When w e attend to one thing , processing of information about un attended
objects are not merely suspended but actively inhibited (Houghton & Tipper, 1994). Macrae, Bodenhausen, and
Milne (1995) ask ed people to watch a videotape of a C hinese wom an. When race was primed, su bjects had better
access to racial stereotype information but access to gender information was inhibited. The reverse was true
1
One of my daughters w ho spent som e time in a fairly remote area of China , and who was quite
tolerant of such unaccustomed attention, did report that she had an almost irresistible urge to scream boo
at small Chinese children who examined her intently and at close range on trains and buses.
2
One of my students recently complained that people starred at her because she was wearing lots of
earrings, nose-rings and other facial hardware. It seems to me, perhaps somewhat naively, that the
purpo se of su ch ad ornm ents is to attract attentio n altho ugh n ot presu mably from st odgy adult s.
3
In fairness, th ese studies d id not ac tually mea sure attentio n, but oth er dependen t measu res were
cons istent w ith the id ea tha t more a ttention had b een give n to th ose in th e visua l line of fire.
Chapter 4: Schemas / Version of November 10, 2002
5
when gender was primed. In a study by Dijksterhuis and van Knippenberg (1996a) when a category label was
primed, subjects had greater access to stereotype consistent and less access to inconsistent information than
people for whom the category was not primed.
Labeling and Interpretation
Attentional mech anisms prov ide a kind of initial gate that allows certain information to b e processed
further, but attention does not guarantee that further processing will take place. If it does, however, the
information must be labeled and interpreted. Human behavior is mostly ambiguous and takes on its meaning
from the context in which it occurs and inferences we make about the intentions underlying it. A smile can
indicate endearment, encouragement, greeting, pleasure, ingratiation, condescension, an overture to sex.
Labeling may be a necessary first step, but we also evaluate new information in terms of it validity and
usefu lness; th ese evalu ations are sch ema d riven.
Evaluation of New Information: Our fast-paced senses and co gnitive processes provide u s with new
information at an alarming rate, and we have to make conscious and unconscious choices about which is useful
and accu rate. That’s wha t our schema ta do for us. Peop le with strong schem ata or theoretical comm itments
evaluate evidence supporting their schemata more favorably than inconsistent information (e.g., Koehler, 1993;
Lord, Ross, & Lepper, 1979; Miller, McHoskey, Bane, & Dowd, 1993; Munroe & Ditto, 1997) and see sources who
disagree w ith them a s biased (Va llone, Ross , & Lepper, 19 85; Giner-So rolla & Ch aiken, 199 4).
Dacher Keltner and Robert Robinson have argued that people view the world from a perspective of naive
realism: they assume that they are in touch with an external reality and that their cognitions, beliefs, attitudes,
and o ther co gnitive s tates reflec t this reality . Therefo re becau se we a ssum e that o ur beliefs a re anch ored in
reality, it must follow that those who disagree are out of touch with reality and thereby biased. As a consequence
we tend to see those who disagree with us as holding extreme positions (e.g., Keltner & Robinson, 1997; Robinson
& Keltner, 1 996; Ro binson , Keltner, Wa rd, & Ro ss, 1995 ), a perception th at mak es our ow n position more
reasonable than our opponents’. In some cases we even see people who agree with us as more extreme than our
own perspectiv e whic h furth er reinforc es the n otion t hat ou r own views a re mod erate an d well-co nsidere d.
Interpretation of Ambiguous Behavior: We all need rather constant reminding that there is little about
behavior, at least at the level most of us are interested in, that is objective. You might describe the gross motor
movements, the blinks, facial flicks and flacks, arm and finger movements, postures fairly objectively in the sense
that another person or an artist could reproduce them. However, such a catalogue would be painstaking and time
consuming even for a brief bit of behavior, and in the final analysis it would tell us almost nothing about what
had happened. Gross bodily movements can mean various things. For example, smiles can be signs of pleasure or
of happiness (which are not, at least in Western culture, quite the same thing), of derision and hostility, of
sociability. To be sure one might point out that the smile of happiness has only superficial features in common
with the smile of derision. And just to make things even more complicated, there are multiple ways of displaying
and indicating almost any emotion or thought. One can be hostile by withdrawing from a situation or by
approaching a hated other, through hitting, stabbing, shooting, by various kinds of well recognized hand
gestures, through a n infinite array of words, by creating a murd erous look. There are relatively few direct,
una mbigu ous c orrespo nden ces betw een beh aviors and s pecific me aning s of tho se beha viors.
If stereotypes are essentially schemata, then we ought to find that they influence the interpretation of
behavior. In on e of the most famo us demon strations, David R osenhan (1973) and seven other people checked
themselves into mental hospitals with vague complaints of hearing voices. The information they provided to the
admitting staff was otherwise truthful, and much to their surprise all were readily admitted with the usual
diagnosis being sc hizophrenia. W hile on the wards they took notes b ut otherwise beha ved norma lly, which is to
say (given that it is hard to behave “no rmally” in such an alien environm ent), they did not further try to simu late
mental illness. Their true status was never uncovered by the mental health professionals (although many of the
other pa tients did sugge st that they w ere fakes), a nd w hen th ey were d ischa rged th ey were g iven th e diagn osis
of schizophrenia in remission. While on the ward the minimal contacts with the professionals led to some
misinte rpretatio ns. Fo r examp le, restless pa cing ou t of bored om w as interp reted as nervo usne ss. On ce one is
labeled o r diagn osed a s men tally ill oth ers can easily in terpret beh avior a s supp orting th at diag nosis .
A classic experiment by Duncan (1976) makes this case for race. He had wh ite subjects watch a videotape
of an encou nter between two m en which tu rned increasingly ho stile until one shoved th e other. Dunca n varied
the race of the two men such that black and white men shoved both blacks and whites. Subjects labeled the
shoving by black stimulus persons as more violent than similar behavior of white stimulus persons. They also
attributed th e behavio r of the black shover m ore to dispo sitional forces than th e behavio r of the wh ite shover.
Chapter 4: Schemas / Version of November 10, 2002
6
Although stereotypic beliefs were not assessed in this study, it is reasonable to expect that white subjects had
beliefs that blacks are more violent than whites, beliefs which, in turn, directed their interpretations of the
behavior. Other studies also show that the behavior of black males is interpreted as more hostile than that of
whites even by black s ubjects (Saga r & Sch ofield, 19 80).
B shoves B
B shoves W
W shoves B
W shoves W
Playing around
0
3
25
19
Dramatizes
6
6
38
44
Aggressive behavior
25
16
25
24
Violent Behavior
69
75
13
12
Table 4.1 Percentage of subjects who coded behavior with each designation.
Another salient stereotype about African American males is superior athletic ability. Stone, Perry, and
Darley (1997) ha d subjects listen to a tape recording of part of a college basketball gam e, and they were ask ed to
evaluate the performance of a player who was either identified as black or white. While subjects’ memories for
actual objective performance (e.g., rebounds) were not affected by the race manipulation, when the player was
black he was perceived to have more basketball ability and to be a better team player. The “white” player, by
contrast, was seen to have more “court-smarts” and “hustle”. In a related study people saw athletic success for
blacks as du e to natural ability and fo r whites due to ha rd work an d better access to coac hing and fac ilities
(Johnso n, Hallina n, & Wes terfield, 1999).
A study by Darley and Gross (1983) makes a similar point about the importance of schemata for social
class. Subjects saw a videotape of a yo ung girl, Hann a, performing various kinds of intellectual problems; her
performance was inconsistent, neither strikingly good or poor. When Hanna was lower class, subjects saw her as
havin g perform ed less w ell on an achiev ement t est tha n wh en she w as iden tified as m iddle cla ss. This study is
especially interesting because when subjects simply were given information about Hanna's social class without
seeing her behavior, their inferences sho wed no effects of socia l class. Perhaps in that kin d of situation su bjects
guarded against showing their class biases. In any event, whatever stereotypes the subjects had about social class
seemed to serve primarily as filters used to interpret Hanna's ambiguous behavior. When behavior is less
ambigu ous, stereo type expectan cies play less of a role (Kam eda, 198 5).
Our expectancies create other effects that are important in this realm. Seta and Seta (1993) argued that
when people violate our stereotypes behaviorally we may assume that they will be more likely to perform some
future stereotypic behavior. It’s as if we make a cognitive effort to balance the behavioral scales in favor of our
hypotheses. So, for exam ple, when subjects read a bout a m inister who performed so me mildly unexpected
behavior (reading a sexually explicit magazine for pleasure), they were more likely to expect him to donate to a
charity (a stereotypically consistent behavior) than when they had not read about such behavior. When he was
seen as havin g perform ing a st rongly unexp ected be havio r (havin g an a ffair with a mar ried wo man and s exually
molesting a child), they did not predict that the minister himself would be more charitable, but they did expect
that a different minister wou ld be. We seem to expect tha t a people in particular groups w ill perform a set
amo unt o f expected behav iors, an d that we ass ume fu ture beh aviors at either th e individ ual or g roup lev el will
comp ensate over tim e.
Gender Biases: Surely g ender a lso affec ts how we interp ret beha vior. A w idely cited study by Ph illip
Goldberg (1968 ) initiated a controversial area of research for perceptions of male and female competence.
Goldberg gave his female subjects scholarly articles to read and evaluate. Some subjects saw that the articles had
been au thored by a w oma n and others by a m an. So me artic les were ev aluat ed mo re highly by thes e female
students w hen they were su pposedly auth ored by male rather tha n female autho rs. Because this stud y seems to
be such a clear dem onstratio n of bias a gainst fem ales in evalu ation of th eir work, a to pic of mu ch con temporary
conc ern, it ha s spaw ned a who le body of researc h.
It is important to be clear as to w hat the original Goldb erg paper did and d id not show since it has been
widely mis-cited. Goldberg gave the subjects six essays supposed authored by male or female scholars in various
areas. For three of the essays th ere were significant differences. The students th ought essay s written on law, city
planning and linguistics were better when "authored" by a male, but for essays in the fields of dietetics,
education, and art history there were no significant differences. Two of the three areas where significant
differences w ere found are tradition ally male area s (law an d city plann ing) and two of th e fields where th ere
Chapter 4: Schemas / Version of November 10, 2002
7
were no differences are traditionally female (dietetics and education). Subjects seemed to be taking additional
inform ation, n amely t he gen der linka ge of the field, into a ccou nt in m aking their dec isions.
A related body of researc h wa s pione ered by P heterso n, Kies ler, and G oldberg (1972 ). They h ad fem ale
subjects evaluate paintings supposedly executed by male or female artists. They also gave subjects information
about whether the paintings had won a prize. This study showed a clear gender bias for ratings of the
competence of the artist with the male being seen as more competent than the female but only when the paintings
had not won a prize; this effect washed out when the paintings were seen to be prize-winning. Many studies,
(e.g., Heilman, Martell, & Simon, 1988) find that gender stereotypes about competence in particular domains can
be over-ridd en by da ta on ac tual perform ance or a bility (See further disc ussion on p. ?).
Several dozen studies have looked at variations on the Goldberg and Pheterson, et al. paradigms. and the
results in this area are surprisingly inconsistent. In a review of the existing published literature, Swim, Borgida,
Maruy ama, an d Myers (1 989) fou nd tha t overall there are g ender biases in such research bu t that the effects a re
quite sm all and basica lly trivial. 4 This does not mean that gender is not important in work-related performance
but only that its effects are likely to be subtle and show up only in certain circumstances.
Explanations and Attributions: Such studies then suggest that at least with ambiguous behavior, our
schemata have an impact on the ways we interpret and label that behavior. The exact mechanism for this has
never been adequately spelled out, but the simplest possibility is what we might call straight-line prejudice. If
some one h as a st rong s tereotyp e that fem ales are in tellectua lly incom petent, h e migh t simply and fa irly
reflexively rate their work as less accomplished. Jussim, Nelson, Manis, and Soffin (1995) argue that labels and
expectancies based on them may arouse affect or emotion which in turn may affect ratings without additional
cogn itive med iation. S o, for exa mple, w hite su bjects m ay sim ply like wh ite actors more th an bla ck on es, and this
greater liking m ay lead th em to label w hite behav ior in more po sitive terms. In s everal experimen ts these au thors
had subjects rate the word definitions provided by rock musicians or child abusers, and as expected the same
definitions provided by th e former were less as more creative an d less pathological tha n those asso ciated with
the latter. However, whereas liking for the two groups did mediate the effects of labels on ratings, beliefs about
the creativity and likely pathology of the two groups did not. This suggests that affect or prejudice rather than
stereoty pic beliefs m ay ha ve been largely res pons ible for the different ev aluat ions o f the stim uli. Ho wever, w hile
the role of affect in interpretation may be impo rtant, it cannot be the total explan ation for schema labeling effects
since th ey are oft en quite specific to stereoty pe relevan t inform ation.
Another possibility is that when a schema is primed it may affect how behavior is interpreted. We may
require less or less strong evidence to judge a behavior in line with the schema (Biernat & Kobrynowicz, 1997). So
we may n ot require extreme behavior or m any instanc es to label behavior as v iolent when w e expect to see
violent behav ior.
Third, ou r schema ta may simply enco urage u s to pay a ttention to s ome featu res of behav ior and ign ore
others. If I believe th at wom en are likely to be in competen t at mech anical tas ks, I may focus on how m uch m ore
slowly a woman works on repairing a car than a man does and ignore the fact that she manages to be less messy
and is less likely to make mistakes than her male counterpart. As a result I may see her as less competent than the
guy even though in some ways she is more so.
Fourth , we may see stereotype co nsistent beh avior as m ore dispos itionally cau sed (Ku lick, 1983).
Wittenbrink, Gist, and H ilton (1997) sugg est that some people ha ve theories about m embers of stereotyped
groups that allow them to interpret behavior in a stereotypic way. Specifically they argued that high and low
prejudiced people have different causal theories about the behavior of African Americans which might affect
interpretations of the behavior. In their research subjects read about a fight between a white and a black man
which resulted in injury. Th e high prejudiced subjects saw the black man as having m ore causal respon sibility
than the white both when he was the victim and the assaulter. Those low in prejudice on the other hand saw the
white as mo re responsible. Similarly in the previously described stu dy by Du ncan (197 6) (see p. ?) where white
subjects watched a black or white male shove another male who was either black or white, subjects attributed the
black shover's beh avior to internal chara cteristics more than to external, wh ile the reverse tended to be true when
the shover wa s white. Apparently these su bjects felt that blacks are especially angry or dispos itionally hostile,
4
How ever, even sm all biases can have larg e comm utative effects (M artell, Lane, & E mrich, 19 96).
Even sma ll biases against a grou p of people at each stage of job a dvancem ent (for example) can add up to
cons iderable re ductio ns in th eir represen tation at high er levels.
Chapter 4: Schemas / Version of November 10, 2002
8
whereas the hostile behavior of whites is brought on by the situation. This same sort of thing was shown in a
field study by Phillips and Dipboye (1989) who found that interviewers who had a pre-existing positive
evalu ation o f an ap plicant rated a good interview perform ance a s more interna l than w as a po or perform ance.
Kay Dea ux and h er colleagues (Deaux, 19 84; Deaux & Emswiller, 1974) ha ve suggested tha t we attribute
the outcomes of male and female performance differently. Specifically, we are inclined to see males as succeeding
because of high ability and failing because of low motivation and females succeeding because of effort and
failing because of low ability.5 A review of the extensive research literature on this topic (Swim & Sanna, 1996)
suggests that for traditionally masculine tasks success is attributed to ability for males and to effort for females,
and that m ales are more likely to be seen as failing for low effort or bad luck. Howev er, these effects were quite
modest in size.
As I suggested in Chapter 3, sometimes categories act as causes of a sort. One Christmas season I found
myself shopping at a Boston department store a couple of evenings before Christmas. The store was crowded, the
clerks were busy and hassled, and the customers seemed tired and cranky. I joined a line of several people to pay
for my purchases, and as the clerk disposed of the customers in front of me, it became clear that she was not your
prototypic helpful salesperson. She growled at people and generally communicated that her job was to ring up
sales and not answer questions. The man in front of me happened to be a black physician (one discovers such
things – the physician part – in long lines), and he ha d seemed pleasan t enough in th e course of our rath er
impersonal con versation. Wh en it became his turn, the clerk sighed deeply as it became clear that this cu stomer
would be the troublesome sort who had questions. He posed his questions, she told him that he should ask
some one on the floor , and h e replied th at he h ad no t been a ble to find a clerk to help him and th at he w ould
appreciate her help. She became more agitated, but he remained surprisingly calm for a minute or two. Finally he
told her that he wa s busy, that he expected her to answer h is questions, and th at if she couldn't he w anted to
speak to th e floor man ager. All this w as said in what I p erceived to be a cool and non-ag gressive wa y. In her turn
she began to tell him that he was being rude and inconsiderate not only to her but to the 10 or so people waiting
in line behind him. He responded by saying (quite reasonably, I felt) that if she had tried to answer his questions
when he first as ked, he w ould h ave bee n ou t of her h air long ago. Sh e respon ded th at she d idn't eve r wan t him in
her hair, and asked whether he wanted to buy his merchandise or not. At this point his voice did get angry, and
he told her that she was one of rudest sales clerks he had ever encountered; he wanted to speak to the floor
man ager. In due co urse, th is person arrived and led our ph ysician friend aw ay so t hey co uld dis cuss his
complaints in a qu ieter place (and away from the other custom ers). She cleared her register, and as I han ded her
my items she said (and I'm quoting nearly exactly)6: "Damned aggressive niggers. They're so touchy and pushy.
5
There is a paradox of sorts in that a sense of achievement and responsibility ought to be enhanced by
effort rather than by ability attributions. Shouldn't we be prouder of the young track runner who works
hard to develop her limited natural ability than the young woman who achieves the same outcomes but
coasts on h er greater talents. Wouldn't you be prouder of a grad e or job promotion th at you ach ieved
through hard work rather than natural ability? Our culture tells us that hard work is important, but at the
same time we reward (especially at high levels of performance) those with natural abilities. Most
abilities, however, are more stable than most dispositions to work hard. If one is born with a beautiful
voice or the ability to run fast and jump high, these abilities are not likely to vary dramatically from day
to day or week to week. Putative math ability which comes and goes is not, we suspect, real ability. Put
another wa y, as individuals w e have little control over how much a bility we have althou gh, of course,
we do h ave con siderable con trol over ou r willingness to display it. Ma tters are different w ith the effort
family, becau se effort is such a matter o f self-discipline and control. So me people are d ispositiona lly hard
workers in the sense that they characteristically work long and hard. But we also know that even our
most habitu al and self-con trolled d esires to ex ert high effort fail u s som etimes. T hus, th ere is a sen se in
which abilities do not desert us but that hard work can and does. That is why it is so damaging to have
one's strong performa nces ascribed to ha rd work as o pposed to ability. Thus , women w ho are perceived
to succeed because of hard work are approved but not as much as their male comrades who are thought
to suc ceed bec ause t hey "h ave th e good s" in term s of ability .
6
Altho ugh th is inciden t happ ened o ver 20 y ears ag o I wro te dow n a det ailed des cription of it with in
an ho ur precis ely beca use it seem ed to be such an inte resting d emon stration of som ething. T he exam ple
Chapter 4: Schemas / Version of November 10, 2002
9
They just want everyone to do everything they want.” I asked the woman no questions and fled with my
purchases a s quickly as possible. 7
What interested me the most about this whole thing was her subsequent interpretation of his reaction as
due in part to his race. I'm not sure how she would have responded to me had I behaved the same. She might
have said: "Some customers are just so hostile", or she might have blamed it on the stresses of the season. But she
would not have accused me of behaving as I might have because of my race, hair color, height, or age. What
interested me most was that she went beyond a mistaken dispositional attribution about the customer being
hostile. She assumed he was hostile to her as a white person because of his race.
Priming: The interpretation of ambiguous behavior can be affected by priming relevant categories. You
may have noticed, for example, that when you leave a violent movie, you are "primed" to see the world in terms
of its violent content. In an initial and influential study, H iggins, Rholes, and Jo nes (1977) exposed subjects to
positive trait words (adventurous, self-confident, independent, persistent) or similar words with negative connotations
(reckless, conceited, aloof, stubborn). Then in an os tensibly unrela ted stu dy su bjects rea d a brief es say ab out D onald
depicting ambiguous behaviors related to the trait adjectives. When they had previously been exposed to the
positive traits, subjects saw Donald's behavior in much more positive terms. There are now dozens of
demonstrations of exactly this kind of priming effect (Higgins, 1996). Interpretations of ambiguous material can
be affected by prim es people h ave rece ntly been exposed to, even when they are una ware o f the prim es.
Factors That Affect Priming: But the effects of priming are often complex. In the first place most of the
early studies produced assimilation-type effects where the prime exerts a cognitive “pull” on the target stimulus,
when the stimulus gets interpreted as close in meaning to the primes. However, sometimes contrast effects occur
when the exem plar is dis placed awa y from t he prim e or wh en prim ing retar ds pro cessing of related stimu li
(Glaser & B anaji, 199 9). Social an d cogn itive psycho logists hav e devoted considera ble energy in recen t years
trying to und erstan d wh at facto rs favo r assim ilation v ersus c ontra st (Higg ins, 19 96; Fo rd & Th omps on, 20 00).
One traditional argument is that assimilation effects are more likely when prime and target are close
together and contrast as they are more distant, and by and large that is true (Herr, 1986; Herr, Sherman, & Fazio,
1983; Manis, Nelson & Shedler, 1988). One way to look at this is that assimilation is most likely when the prime
and th e target o verlap in mean ing or ev aluat ion. If th e stimu lus is am biguo us eno ugh to require in terpretatio n in
terms of the prime (Stapel & Schwarz, 1998), or if the primed categories are sufficiently broad so as to include the
stimulus (Stapel & Koomen, 2000) assimilation will occur. In a related argument Schwarz and Bless (1992a) argue
that assimilation w ill occur only if the prime is part of the larger target category. So if people are primed w ith
Richard Nixon, they might judge that the larger category of politicians (which includes Nixon) are less
trustworthy. However, a Richard Nixon prime might lead people to judge Bill Clinton as more trustworthy since
Nixon is n ot part of the C linton categ ory. Similarly, exa mples of A frican-Am erican su ccess prod uce mo re
negativ e stereoty pes of M exican -Americ ans (H o, San bonm atsu, & Akimo to, 200 2).
However, when the prime and target do not overlap, the prime is more likely to be seen as a standard for
compa rison an d if sufficiently distin ct to lead to c ontrast. So , for example, distin ct primes typica lly have a m ore
narrow range of interpretative possibilities than category primes and thus present themselves as more salient
standa rds of com parison ra ther than for interpretation (Stapel & K oomen , 1997a ). Also wh en conc rete exemplars
are used as a prime, there is no guarantee that features relevant to the target will be activated as a part of the
interpretive enterprise (P hilippot, Schw arz, Carrera , De Vries, & va n Yperen , 1991). Th us, con crete exemplars are
generally should usually lead to contrast effects unless the concrete prime is very close to the stimulus
(Dijksterhuis, S pears, & Lép inasse, 20 01; Wä nke, Bless, & I gou, 20 01).
Stage in processing may also make a difference. When the prime acts as an interpretative schema at time
of the perception and encoding of the target assimilation is likely, whereas at a later, more reflective phase, the
prime is seen as a standard of comparison contrast is more likely. Using the prime as a standard of comparison
has fin ally fou nd a u seful ho me.
7
Nowadays I would be more inclined to have confronted the woman about her racist comments, at
least enough to let her know that I did not approve of her behavior, even though it would likely not
affect her stereotypes. But I was tired, in a hurry, and frankly a bit scared about what I had just
witnessed.
Chapter 4: Schemas / Version of November 10, 2002
10
requires a co nsciou s comp arison, an d this can not occ ur when the person is unaw are of the prim e. Therefore
when people are relatively una ware of primes, assimilation resu lt.
Often contrast effects are found when the individual is aware of the prime (Lombardi, Higgins, & Bargh,
1987; N ewma n & Ulem an, 199 0; Strack, Sc hwarz , Bless, Küb ler, & Wän ke, 1993 ), because th e person is m ore
likely to see the prime as a comparison because the conscious prime gives rise to more thoughtful processing.
When a comparison is salient, people may be awa re of its “pull” and often attem pt to correct for what they
imagine to be the biases it produces (Wegener & Petty, 1997); this may produce overcorrection and consequent
contras t. Martin (19 86; Ma rtin & Ach ee, 1992; W egner & P etty, 1997 ) has su ggested tha t contras t requires mo re
cogn itive effort a nd is les s likely to o ccur w hen th e individ ual is eith er not m otivat ed to o r cann ot explicitly
comp are the p rime an d the st imulu s to be ju dged (M artin, Set a, & Cr elia, 199 0). As w ould b e expected from th is
perspective, primed categories produce contrast when people have full use of their cognitive capacities, but
assimilation results with high cognitive loads (Ford & Kruglanski, 1995; Newman, Duff, Hedberg, & Blitzstein,
1996). Motivational factors such as a desire to be accurate (Sedikides, 1990; Stapel, Koomen, & Zeelenberg, 1998;
Thom pson , Rom an, M osko witz, C haiken , & Barg h, 199 4) also reduc e assim ilation, b ut wh en interp retive go als
are salient as similation is m ore likely (Stapel & K oomen , 2001).
Priming and Stereotypes: In everyda y life categories cen tal to stereotyp ing such as race an d gender a re
frequently primed through the mass media, humor, and conversation. These primes are likely to be somewhat
subtle and beca use they are relatively non-con scious ma y give rise to assimilative processing. For exam ple,
Dovidio, Evans, and Tyler (1986) have shown that priming of race categories can facilitate the availability of race
related terms; this is particularly strong for prejudiced subjects (Wittenbrink, Judd, & Park, 1997) and for those
who think egalitarian values are no t especially important (Mosk owitz, Salom on, & Taylor, 20 00). Such primes
may affect interpretation directly. Wittenbrink a nd Henly (19 96) primed racial stereotypes by a sking subjects
questio ns des igned to elicit high or low estima tes of bla ck stereo typic beh aviors . Thos e subjec ts wh o were h igh in
prejudice were more likely to see a black defendant as guilty in a mock trial when their stereotypes had been so
primed. Whites primed w ith black faces identified pictures of gun s as guns m ore rapidly than w hen primed with
white fa ces (Pa yne, 20 01).
Gender stereotypes can also be primed. Bana ji and Hardin (19 96) used a primin g manipulation to
demonstrate automatic effects of gender stereotype activation. Subjects were exposed to brief presentations (200
ms) of gen der related w ords (e.g., nu rse, father, Mr., sales man, etc.) a nd then after a brief delay (1 00 ms) a
pronoun was presented, and the subject was asked to indicate whether the pronoun was masculine (e.g., his) or
feminine (e.g., her). The basic idea is that when subjects have been primed with a feminine noun, they will be
able to judge the feminine pronouns more quickly than when they have been primed with masculine nouns.
Similarly, masculine nouns should prime judgments of masculine pronouns. Gender consistent noun-pronoun
pairings did lead to faster judgments than inconsistent. These results support the notion of automatic activation
of gend er stereoty pes beca use giv en the b rief time inte rvals, su bjects co uld no t have contro lled their res pons es in
any m eaningfu l way. At a minimu m then, th ese results su ggest that w hen peop le see words such a s nurse, fath er,
salesman, m echanic they au tomatically tend to th ink of that person in terms of gender
Priming effects may also occur throu gh judging the beh avior of a category m ember. So when whites see a
negative behav ior by one black m ale or even hear abo ut a crime com mitted by a black, this m ay lead to increased
stereotyping of blacks as w ell as increased white in-group fa voritism (Henderso n-King & N isbett, 1996). When
people infer traits from behaviors for black m en, stereotypic trait judgments for other blac k men are facilitated
(Stewart, Doan, Gingrich, & Smith, 1998). Judging familiar people facilitates judgments of unfamiliar people of
the same age and gender (Karylowski, Konarzewski, & Motes, 2000)
In our everyday interactions, priming may result from slurs or humor. Greenberg and Pyszczynski (1985)
found that subjects who overheard an ethnic slur (presumably priming a racial category) interpreted the poor
performance of a black target more negatively, and an ethnic slur directed at an attorney lowered his general
ratings (Kirkland, Greenb erg, & Pyszczy nski, 1987). Further research by Simon a nd Greenberg (199 6) showed
that ethnic labels affected evalua tions only for sub jects who had negative attitudes tow ard blacks; those w ith
more a mbiva lent attitu des resp onde d to eth nic slu rs by ev aluat ing the b lack ta rget mo re positiv ely.
What ab out ethnic hu mor? Ha rmless enough , isn’t it? Ford (1997) foun d that after watch ing racestereoty pic hu mor, w hite su bjects rat ed a bla ck defen dant as mo re guilty th an afte r watc hing n on-ster eotypic
humo r. Not so harm less apparently. How ever, effects for other categories may be different. In several stud ies
(Olson, Maio, & Hobden, 1999) jokes about men and lawyers had essentially no effects on stereotyping thinking.
Ford (2000) a rgues that hu mor often leads to a non-critical approach to stereotype related material, and tha t when
Chapter 4: Schemas / Version of November 10, 2002
11
people are non-critical in this way exposure to sexist humor leads to greater tolerance of sexism especially for
those high in hostile sexism.
There are also demo nstrations that prim ing can affect stereotypic processing of information. For exam ple,
when tradition al gend er stereoty pes are pr imed, su bjects are more w illing to in terpret the b ehav ior of w omen in
terms of that stereotype (Hansen & Hansen, 1988). Men who have a strong schema for gender respond to sexual
arousal by emphasizing the sexual characteristics of women they meet (McKenzie-Mohr & Zanna, 1990). Violent
rap music (which is associated with African-Americans) increases dispositional judgments for a black’s violent
behav ior but h as no effect on similar b ehav ior by w hites (Joh nson , Traw alter, & D ovidio , 2000 ). Sexua l rap mu sic
(which is often deroga tory of wom en) produces m ore negative evaluation s of black wom en by white sub jects
(Gan, Zillmann, & Mitrook, 1997). In a striking demonstration of priming traditional sexual stereotypes, Rudman
and Borgid a (1995) ha d subjects watc h actual television ad vertisements that w ere either sexist or non-sexist.
Following this, they then rated a female job applicant whom they interviewed. Subjects primed with the sexist
humo r asked more sexist ques tions during the interview , recalled more information a bout the applican t’s
physical features and clothing, and behaved in a more sexualized way during the interview at least as rated by
the applicant. How ever, the primed subjects were also m ore willing to hire the applicant but n ot because of her
qualifica tions b ecaus e they ra ted her lo wer on comp etence b ut high er on frien dliness .
Most of these studies involve the effects of priming on the perceptions and evaluations of others, but
under so me con ditions prim ing can a lso affect ou r own b ehavior. D ijksterhuis an d Bargh (2001) a rgue that th ere
is an automatic perception-behavior link -- merely observing behavior increases the likelihood of performing that
behavior. Certainly there is evidence that observing the motor behavior of others leads to non-conscious imitation
and that when one participant mimics the behavior of the other, the interaction goes more smoothly (Chartrand
& Bar gh, 19 99).
There are several demonstrations that priming of stereotype categories can lead to stereotype consistent
behavior. In a clever study Becca Levy (1996) primed elderly subjects with either senility or wisdom, two
contrasting aspects of the stereotype of the old. Those who were primed with the term wisdom actually had
better memory scores than those who were primed with senility, and the former also rated the ambiguous
behavior of an older woma n more positively. Elderly people primed w ith negative stereotypes assoc iated with
old age were less likely to endorse the use of life-prolonging medical treatments in hypothetical situations than
those primed with relatively m ore positive stereotypes (Levy, Ashm an, & Dror, 200 0). Such priming even affects
meas ures of c ardiov ascu lar respo nses to threat (L evy, H ausd orff, Hen cke, & W ei, 2000 ).
Such effects are not restricted to those w ho are stereo typed. For exa mple, when college age stu dents are
primed with elderly stereotypes, they perform less well on a memory test but only if they report frequent contact
with the eld erly, the latter creating a stronge r associa tion betw een the elderly ca tegory an d stereotypic m emory
impairm ent (Dijks terhuis , Aarts, B argh, & van K nipperb erg, 200 0). In a conc eptually related st udy, D ijksterhu is
and v an K nippen berg (19 98) sh owed that w hen su bjects w ere primed with th e profess or role or the trait
intelligent, they performed better on a general kno wledge test but performed w orse when primed with soccer
hooligan or stupid.
Bargh, Chen & Burrows (1996) showed priming effects for several different more overt behaviors.
Subjects primed with rudeness were more likely to interrupt and experimenter than those with no prime or who
were primed with polite. A black face primes hostile behavior more than a white face does. When college age
students are primed w ith age stereotypes, they walk m ore slowly. Earlier I suggested tha t exemplars more often
lead to contrast than do trait primes in interpretative tasks, and the same appears to be true for performance
measures. Su bjects primed with professor stereotypes perform ed better on a know ledge test than those primed
with the su permod el stereotype, but th ose primed with extrem e exemplars (E instein an d Clau dia Schiffer)
show ed con trast effect s (Dijkster huis, S pears, P ostm es, Stap el, Koo men, v an K nipperb erg, & Sc heepers, 1 998).
Stereotype Threat: The Levy research cited a bove sugges ts that people who are victims of stereotypes
may behave consistently with those stereotypes. Claude Steele (1997) has pointed to a related phenomenon he
calls stereotype threat. 8 When m embers of stereotyped grou ps are aware of nega tive stereotypes about them , they
8
While stereotype threat is similar in som e ways to ass imilative priming, generally they have been
discussed in quite different ways (see Wheeler & Petty, 2001, for extended discussion). One obvious
difference is that whereas stereotype threat focu ses on decremen ts in performance, priming ca n facilitate
as well as hinder performance. Also while priming is usually discussed as a fairly automatic process that
Chapter 4: Schemas / Version of November 10, 2002
12
may fear that others will apply these negative qualities to them. These thoughts may be anxiety provoking and
lead to lower performance on a variety of tasks. Steele points to the fact that although standardized tests predict
black and white performance approximately equally well, at every level of intellectual ability and preparation,
black students tend to perform less well. Perhaps this is because of the performance anxiety they experience
because of stereotype threat. 9 Women also “underperform” (relative to performance predicted from ability tests)
in math and science courses arguably because of threat from the stereotype that women are not clever in these
areas. It’s important to realize that these effects do not apply to everyone in negatively stereotyped groups. For
stereotype threat to do its dirty work, people from stigmatized groups have to be identified with the area of
performance and see performance as linked to the group (Schmader, 2002). My jumping ability is not likely to be
affected if I don’t think wh ite men can’t jump o r that I basically think jum ping is irrelevant. By the time they get
to high school many minority students may have decided that school is more or less irrelevant to their lives and
young women may ha ve abandoned math and science – the stereotype threat notion does not apply to these
students, but o nly to those wh o still care about performing w ell. Finally it is probable that stereotype threat does
not affect perfo rmanc e on extremely easy or imp ossibly difficu lt tasks, but m ainly on th ose wh ere abilities are
stretche d to th eir outer lim its.
Confirmation for this hypothesis was found in a number of studies by Steele and Aronson (1995). Black
and white students, matched in ability, were asked to take difficult intellectual tests (mostly subsets of the
Graduate R ecord Exam ). When the test wa s presented as a diag nostic test of intellectual ability black students
performed worse than whites although the two groups performed equally well when the test did not have these
diagno stic overton es. The blac k studen ts tended to be awa re of black stereo types wh en they tho ught th ey were
being tested o n intellectual a bilities, and presu mably th is led to stereotyp e threat. Mo reover, the effects w ere
stronger for black stud ents when th ey were identified by race suggesting th at stereotype threat is stronger wh en
race is salient.
People from lower social classes who stereotypically perform less well on intellectual tasks are also
vulnerable to stereotype threat (Cro izet & Claire, 1998). Similar effects have been fou nd for wom en taking ma th
cours es (Spen cer, Steele, & Quin n, 199 9). Altho ugh w omen and m en perfor m equ ally well o n mo derately
difficult math tests, the men ou t-perform the wom en on tests of greater difficulty. This difference was preserv ed
when the test was described as one that typically produced gender differences, but more importantly it was
eliminated when subjects were told that the test was gender neutral, thus presumably reducing stereotype threat
to the women. While women seem most vulnerable to stereotype threat when the test is described as one that
measures m ath weakn ess, men show declines in performance w hen the test is described as on e that measu res
exceptio nal m ath ab ilities (Brow n & Jos ephs, 1 999). F emales are mo re vuln erable to stereoty pe threa t abou t their
math abilities when they are in the presence of men, presumably because this makes the math-deficit stereotype
more s alient (In zlicht & Ben-Z eev, 200 0).
As we saw in Chapter 4 som etimes activating a pos itive stereotype can improve performa nce. In a clever
study using Asian-American women and math tests Shih, Pittinsky, and Ambady (2000) found that activating
gender identity led to worse perform ance presum ably because w omen are threaten ed by the stereotype that they
are no t good at ma th. Ho wever, w hen th eir ethnic identity w as mo re salient, th ey perform ed better p resum ably
“living” up to the stereotype that Asians are gifted at math. Other research (Cheryan & Bodenhausen, 2000) has
suggested that when Asian identity is manipulated in a more public way stereotype threat effects are obtained,
perhaps becau se public standard s create a “chokin g under pressure” scenario whereas more private identity
salienc e may motiv ate high er perform ance. O f cours e, math and in tellectua l perform ance a nd no t the on ly
features linked to gender. Stereotypically women are not only relatively poor at math but also at negotiating (and
directly a ffects beh avior, s tereotyp e threat re lies on m ore com plex interv ening m echan isms.
9
Another possible, if controversial, reason for lower achievement performance is the oppositional
identity pressures experiences by many minority students in which high achievement is seen as negative
and is identified by their peers with “acting white” (Fordham & Ogbu, 1986). Furthermore those who
study stereotype threat do not deny tha t there may be other, even m ore powerful, reasons for low er
performance by members of threaten groups – they may, for example, be victims of direct discrimination
or ove rt sugg estions that th ey are no t up to p ar on re levant tasks.
Chapter 4: Schemas / Version of November 10, 2002
13
other forms of assertive behavior), and true to form women are less effective at negotiating when they think
performan ce reflects abilities or is tied to gender-spec ific traits (Kray, Th ompso n, & Galin sky, 200 1).
Stereotype threat need not be restricted to historically stereotyped and stigmatized groups. Aronson,
Lustina, Goo d, Keough , Steele, and Brown (19 99) found th at white male college stud ents who h ad high m ath
abilities (mean above 700 on the math SAT) performed less well on a difficult math test when they were told that
Asian students typically out-performed whites on such tests. However, in a second study this effect was found
only for those high scoring students who said that math ability was important to their self-concept, confirming
the idea that identification with the threatened area is an essential part of the psychological mix for the activation
of stereo type thre at. Wh ite males perform ed less w ell on a g olf task when perform ance w as said to be dia gnos tic
of natural athletic ability whereas black males performed less well when the same task was framed as a test of
sports intelligence (Stone, Lynch, Sjomeling, & Darley, 1999). Males who are told they are poorer at affective
tasks a ctually do perfo rm wo rse (Leyen s, Desert , Croize t, & Da rcis, 200 0).
This idea of stereotype threat is an important one. It is one thing for us to suggest as we did earlier that
knowledge o f others’ stereotypes can som etimes lead to attempts to d isconfirm the stereotype. It is quite another
to assume that such efforts will necessarily be successful or that the person has the intellectual and personal
resources to do the job. Indeed stereotype threat seems to undermine whatever confidence people bring to tasks
(Stangor, Carr, & K iang, 1998). Stereotype threat creates a dded burd ens, in this case performan ce anxiety 10, that
may ma ke it difficult, even impossible, to disconfirm the stereotype bas ed hypotheses o thers have.
Memory
Schemata Aid Encoding: We often find ourselves u sing our schem ata to process inform ation as it comes
to us. But m any of the mo st important judgm ents we mak e in everyday life are based on o ur memories. Sch emata
affect bo th enco ding in to and retrieval o f inform ation fro m mem ory. Ma ny of u s hav e the experie nce of n ot really
understanding what we are reading or hearing when we are exposed to new material in an unfamiliar area, but
finding it much clearer once we get the "big picture". Generally we comprehend and remember relevant material
better when a relevant sc hema is salient (And erson & Pichert, 197 8; Dooling & L achman , 1971). Several studies
(e.g., And erson & Pichert , 1978 ; Stillwell & B aum eister, 199 7) hav e show n tha t the persp ective th at one takes in
readin g mat erial ma y also a ffect wh at one remem bers; a rea l estate ag ent an d a po tential bu rglar w ould s urely
remem ber differen t things abou t a hou se they h ad jus t cased out. A perceiver w ho th inks ab out a wom an in
terms o f her job qu alificatio ns will rem ember d ifferent thin gs abo ut her th an on e who is interested primarily in
her ho mema king sk ills or poten tial as a sexua l partner. O r when a perceiv er think s of an Asian wom an as female
she m ay rem ember lo wer SA T ma th sco res than if she thin ks of h er as As ian (Pitt insky, S hih, & A mbad y, 200 0).
Memory Errors: There is little doubt that relevant schemata facilitate understanding and encoding of
new information in ways that aid later memory. But there is more to memory than simply how much one can
dredge up, and this more is highly relevant to stereotyping. One straightforward prediction from schema
models is that relevant schemata should facilitate schema relevant material more than schema irrelevant or
inconsistent. Our schemata do not like being molested by inconsistent data, and they preserve themselves from
being invalidated. A second important prediction is that sometimes our memories play us false and we
misremember w hat we hav e seen or heard; our sch emata and stereotypes sometimes create false or biased
memo ries that suppo rt their va lidity.
Thus a clear prediction from schema theories is that errors in memory should support the schema. In
particu lar, people a re inclined to remem ber, falsely , that sc hema relevan t materia l has be en presen ted wh en it
has not been (e.g., Bower, Black, & Turner, 1979; Lenton, Blair, & Hastie, 2001; Sherman & Bessenoff, 1999). For
example, people who have strong gender stereotypes are biased to report having seen gender consistent
10
In fairness, attempts to tie stereotype threat directly to anxiety have largely been unsuccessful. Some
studies find that m easured anxiety pa rtially mediates the effect (Osborne, 200 1) and other stu dies
(Blasovich, Spencer, Q uinn, & Steele, 2001) find ph ysiological indicators of stress ac compan y perceived
stereotype threat. Intrusive thou ghts abou t performance ma y be the culprit as Cherya n and Bo denhau sen
(2000) found that stereotype threat lowered concentration. Yet another possibility is that stereotype threat
may in terfere with abilities to formu late an d use p roblem solvin g strateg ies (Qu inn & S pencer, 2 001). It is
also possible that feeling under the gun directly reduces motivation to do well. In fact there are a great
man y explan ations involv ing mo tivation , anxiety , and c ognitiv e proces ses, bu t to dat e there is little
suppo rting evid ence fav oring a ny on e of them (see Wh eeler & Pet ty, 200 1).
Chapter 4: Schemas / Version of November 10, 2002
14
inform ation a bout people (St ango r, 1988 ), and c hildren will often misrem ember th e gende r of acto rs in scen es in
involving gender-incongruent activities (e.g., Signorella & Liben, 1984). More formally, we show a relative
inability to disc riminate presen ted from n on-presen ted materia l that is relevan t to a genera l knowled ge structu re
or stereotype (Graesser, Woll, Kowalski, & Smith, 1980; Lui & Brewer, 1983; Rothbart, Evans, & Fulero, 1979;
Woll & Graesser, 1982). Thus, when our stereotypes are activated, we sometimes have trouble remembering
which of many possible stereotypic behaviors the person did or did not perform. During my annual physical last
year did my ph ysician take m y bloo d pressu re? My t empera ture? A sk me h ow m uch c offee I drin k a da y? Tell
me I need to lose som e weight? Surely som e of these things, but wh ich I really can’t remember for su re.
Reconstructive Memory? Most theorists have assumed that knowledge structures play their greatest
role at time of encoding of new information, at a time when incoming information can be clarified by schemata or
where schemata can facilitate links to other information. However, Bartlett (1932) and others have argued that
memory is essentially reconstructive and that schemata play their largest role at time of retrieval. Extreme
versions of this theory (which are surely false) suggest that we store relatively little information about details of
people and events b ut instead recons truct such deta ils when we need them during reca ll. So, for example, when
one sees a professor behaving in stereotypically professor like ways (driving an old Volvo and taking liberal
political positions), one need not record that individual fact. Rather, one could simply record that this person was
a typical professor and then "retrieve" her stereotypic consistent behaviors at some later point by simply knowing
that she was like most other professors.
One interesting implication of this perspective for reconstructive memory is that it should make no
differenc e wheth er stereoty pes wer e instan tiated be fore the in comin g inform ation o r afterwa rds. If th e main
function of sch emata is to act as a retrieval framework or pattern for reinterpretation, introdu cing a schem a after
presentation of relevant inform ation shou ld be at least as effective, if not more so, than presen ting it before.
Suppose you observe Harry Jon es attending a rally of som e liberal group. You mo mentarily store awa y that fact.
Subsequently you learn that he is a professor. Because you think that professors are politically liberal, that piece
of information should be a good retrieval cue when you are trying to recall information about him despite the
fact that yo u learned it a fter seeing the beh avior. Th ere is some su pport for the reco nstructiv e memo ry
perspective. For example, there have been a few reports of presentation of group information after behavior
affecting memory for beh aviors (e.g., Cohen, 198 1; van Kn ippenberg & Dijksterhuis, 199 6), although stron ger
effects are foun d wh en the in forma tion ab out ca tegory c omes before th e behav ioral info rmatio n and presum ably
aids in its enco ding (R othba rt, Evan s, & Fu lero, 197 9).
Clearly th e recon structiv e mem ory ap proac h predic ts not o nly tha t impos ing a sc hema after the fa ct will
aid memory but it should also produce errors, in particular remembering events or behaviors that are consistent
with the schema but did not actually occur. This has been demonstrated in studies showing that what we do
with information affects subsequent reca ll. For example, Higgins and M cCann (1 984) had s ubjects comm unicate
information about a person to others who were known to have a particular bias for or against the person. As
expected the subjects tended to tailor their messages to their audience. Subsequently measures of recall for the
inform ation s how ed tha t subjec ts disto rted it to b e cons istent w ith their co mmu nicatio n.
Ross (1989) argues that in our memories for our own lives, we often reconstruct the past to be consistent
with the present. For exam ple, when people chan ge their attitudes toward a persuasive com munication , they
recall past behaviors to be m ore consistent with th e new attitudes (Ro ss, McFarland , & Fletcher, 1981). Older
adults who believe that certain traits decline with age, recall their earlier standing on that trait as higher than a
group of younger people, and when they expect improvement, they recall their earlier lives as less good than the
younger adults do (McFarland, Ross, & Giltrow, 1992); in both cases recall of the past supports their theories of
change. Co nway a nd Ross (1984) had subjects participate in a study s kills program. Althoug h their actual grades
did not improve, they recalled the evaluations of their pre-class performance as being worse than they were thus
supporting their sense that they had improved. In a demonstration more relevant to stereotyping, Hirt (1990) and
Hirt, Erickson, and McDonald (1993) investigated whether the same results would occur when we perceive
others. They foun d that wh en subjects expected a stud ent’s performance to im prove they recalled his earlier
performance as w orse than it had been, and wh en they expected decline, they tended to see the earlier
performance as having been higher than a no-expectancy control group. Expectancies are especially important
when they match the subjects’ own liking for the target (McDonald & Hirt, 1997). That is, these reconstructions
are especially likely for those who liked the target and expected improvement and for those who disliked the
target a nd exp ected d ecline.
Chapter 4: Schemas / Version of November 10, 2002
15
Remembering Schema-Consistent vs. Schema-Inconsistent Information: Most schem a mo dels
strongly suggest that there should be strong preferences for schema consistent material in memory. You know
that Janice is a Professor o f Psychology . Clearly that should h elp you remember tha t spends part of her da y in her
lab and has a good knowledge of statistics. It is, however, not
likely to help remembering whether she lives in a brick house
or grew up in Iowa. But what about finding out that she enjoys
stock car races or that spends he free time reading trashy
romance novels.? Is schema inconsistent material is totally lost
in the dense undergrowth of our minds?
Work on this questio n wa s stimu lated by classic
experiments reported by Hastie and Kumar (1979). They found
that schema-relevant information which was both consistent
and inconsistent with the schema was remembered better than
informatio n which was irreleva nt to the sc hema; m ore
striking ly, the inc onsis tent info rmatio n wa s actu ally
remembered better than th e consistent. People also tend to
have especially good memo ries for info rmatio n tha t is
Figure 4.1: Relations of consistent and
inconsistent with behavioral scripts (Bower, et al., 1979), for
inconsistent attribu tes to category
faces (Light, Kayra-Stuart, & Hollander, 1979) and visual
scenes (Friedm an, 19 79; Pez ek, Wh etston e, Reyn olds, A skari, &
Doug herty, 198 9). Such results are no t always found (e.g., Stangor, 1 988; Sta ngor & R uble, 198 9; van K nippenberg
& van K nippenberg, 1994), bu t there is little doubt of the existence of a memory preference for inco nsistent rather
than consistent information under at least some circumstances (Fyock & Stangor, 1994; Rojahn & Pettigrew, 1992;
Stang or & M cMillan , 1992 ).
Theoretical Implications: This preference for incongruent material has important theoretical and practical
implications. At the level of theory, simple schem a models can not easily accou nt for the results 11 because they
emphasize the idea that elements of the schema help one remember information that fits well with the schema.
Hastie (1980) and Srull (1981) have argued that such inconsistency effects are better understood within an
associative network model of memory. During encoding incongruent information receives additional processing
as people try to figure out what it means, and this leads to its being linked to other items of information.
Congruent information slides in without effort and therefore does not get as readily linked to other items of
information. Thus during recall the inconsistent items will have more links with other items which can serve as
addition al retrieval cues , althoug h such retrieval cues a re less relevant for rec ognition (See Figure 4.1).
In supp ort of this m odel, eviden ce sugges ts that preferenc es for the inco ngruen t informatio n are
enhanced when the person has an explicit set to try to organize or explain the incoming information rather than
remember it (e.g., Garcia-Marques & Hamilton, 19 96; Srull, 1981). Interestingly people wh o are highly prejudiced
are especially pro ne to proces s stereotype inc onsisten t informatio n throu ghly and to be threaten ed by it (Förster,
Higgins, & Strack, 2000). We know that people spend more time processing inconsistent information (Stern,
Marrs, Millar, & Cole, 1984) which is what you would expect if people are trying to understand the incongruent
information and linking it to other schema relevant information. Memory preferences for inconsistent
information seem most robust when memory is measured with recall (which takes advantage of the associations
among items) rather than recogn ition (which does not). Also given assu mptions abo ut more links between
incongruent and other items, recall of a congruent item is more likely to be followed by recall of an incongruent
than anoth er cong ruent o ne (Sru ll, 1981 ; Ham ilton, Dr iscoll, & W orth, 19 89) w hich w ould b e expected if
incon gruen t items h ave m ore links to con gruen t items th an do cong ruent item s.
Since the memory preference for incongruent items depends to some extent on the extra processing such
items receive, it follows that when people are unable or unwilling to give extra processing time to such items
11
How ever, schem a mod els with mo re sophistica ted and complex a ssum ptions (e.g.,Bab ey, Queller,
& Klein, 1998; Graesser, Woll, Kowalski, & Smith, 1980; Woll & Graesser, 1982) can explain such results.
Generally this class of models assumes that while processing of congruent information is schematic, the
incongruent information gets special processing and has a "tag" associated with it that sets it apart as not
a part of the schema.
Chapter 4: Schemas / Version of November 10, 2002
16
memory preferences for inconsistent information should be reduced or reversed (Dijksterhuis & van
Knipperberg, 1995b; Garcia-Marques, Hamilton, & Maddox, 2002; Macrae, Hewstone, & Griffiths, 1993; Pendry &
Macra e, 1999; Sta ngor & D uan, 19 91; Wy er & Martin , 1986). 12 When people h ave to reconc ile multiple t rait
expectancies or trait and category based stereotype expectancies, recall preference for inconsistent items is also
reduced or reversed presumably because of increased task demands (Driscoll & Gingrich, 1997; Gingrich,
Hou ston, & L upfer, 1997 ; Hamilton , Driscoll, & W orth, 198 9). People w ho are hig h in need fo r cognitive stru cture
are less likely to show inconsistent information preference (Crawford & Skowronski, 1998; Dijksterhuis, van
Knippenberg, Kruglanski, & Schaper, 1996). Another individual difference measure that affects result is the
implicit theories that people have about traits (See Chapter 14). Those who have entity theories (traits are fixed)
attend to and remember consistent information better, whereas those with incremental theories about traits show
the sam e biases tow ard inco nsistent info rmation (Plaks, Stroes sner, Dw eck, & Sherm an, 200 1).
Practical Implications: Research is this area has been prolific primarily because of theoretical issues, but
there are a lso prac tical imp lication s. If in pro cessing inform ation a bout memb ers of stere otyped group s people
tend to remem ber a dis propor tionat e amo unt o f inform ation w hich d iscon firms th e stereoty pe, then it w ould
seem to follow that stereotypes would soon dissolve under the weight of such disconfirmations. Yet, common
experienc e sugg ests tha t this is ra re. How is it possib le for inco nsisten t inform ation b oth to have a n adv anta ge in
memory and yet to have less impact on our impressions of others? One possible reason is that over time the
discrepant information is forgotten more rapidly than the consistent (Graesser, et al, 1980). Because the amount of
explicit dis confirm ing info rmatio n ma y often b e muc h sma ller than the am oun t of am biguo us or c learly
confirmatory information the disconfirming evidence may lose some of its saliency over time (Rothbart, 1981). Or
perhaps after all the extra processing that discrepant information receives, it is absorbed into the general schema
in a way which alters its meaning or makes it less salient. Not only can inconsistent information sometimes be
reinterpreted to fit the existing schema or stereotype, but it may be seen as less dispositiona l and more su bject to
situational forces (Von k, 1994). One v ivid example (see Schneider, 198 8, for a fuller account) of this occ urred
when a student tried to convince me that Hispanics were actually quite lazy. His evidence was that MexicanAmerican w orkers on his father’s farm took frequent wo rk breaks, but wh en I asked him if they actually picked
as ma ny cro ps as th e Anglo work ers he ad mitted th at they did bu t only b ecaus e they w ere being c arefully
monitored by the field foreman. Apparently even hard work can be seen as a sign of laziness.
Interestingly in a serial reproduction paradigm (where one person hears the story, relays it to a second,
and so forth), people w ho hea r the original n arrative recall m ore incon sistent inform ation as expected. Ho wever,
by the third person in the chain this advantage has washed out and by the fifth the story emphasizes consistent
material (Kashim a, 2000). There cou ld be many reaso ns for this, but one likely one is that a s the story gets
condensed, the inconsistent information gets “cut loose” from the main threads, and is thereby easily forgotten.
Since much of the information we get about others comes in the form of second-hand accounts and even stories,
this may suggest that consensus stereotypes emphasize consistent information even if individually held ones do
not alw ays.
Another poss ibility is that the person displaying the discrepan t information is treated as a n exception to
the stereo type wh ich allo ws th e stereoty pe to be free from co gnitive m olestatio n (Web er & Cro cker, 19 83). It is
also important to n ote that memo ry preferences for inconsistent informa tion occur on ly when people are allow ed
to process information fairly throughly, and in everyday life we may usually have neither the inclination nor
cognitive resources to think about every behavior of every person we meet. Thus, real world processing
conditions may favor consistent information (Hilton & von Hippel, 1996). Finally we note that inconsistency
12
There are some qualifications. For one thing only tasks that interfere with the executive functions
implicated in planning thought and behavior result in eliminating the preference for inconsistent items
(Macrae, B odenh ausen , Schloersch eidt & Milne, 1 999). Jeff Sherm an an d his colleag ues hav e argued fo r a
complex relationships between cognitive resources and differential processing. In what they call the
encoding flexibility model (Sherma n, Lee, Bessenoff, & Frost, 199 8). The basic argum ent is that when
stereoty pes are pr imed, peo ple need t o pay little attentio n to co nsisten t inform ation a nd w ill use va luable
resources for attention to and perceptual encoding of inconsistent items. However, the conceptual
mean ings of cons istent item s are still en coded which gives th em an adva ntage in recall bu t not in
recognition (Sherman & Frost, 2000) because the inconsistent items are individuated (because of
perceptu al enco ding) b ut no t integra ted into the stereo type wh ich resu lts in poo rer recall.
Chapter 4: Schemas / Version of November 10, 2002
17
effects ten d to be s tronge st for info rmatio n abo ut ind ividua ls and weak or revers ed for th at abo ut grou ps (e.g.,
Stern, et al., 1984), so that stereotypes as features associated with groups may be less vulnerable to this sort of
problem. How ever, there is also evidence that wh en group labels are applied to ind ividuals, especially from outgroups, the mem ory preference for inconsistent informa tion is increased (Barda ch & Park, 19 96). Also we tend to
focus more on consistencies for out-groups and inconsistencies for in-groups (Koomen & Dijker, 1997)
presum ably be caus e we are m ore inclin ed to fo cus o n the w ays ou t-grou p mem bers are s imilar (see C hapter 7 ).
Remembering discrepant information is the way we have of individuating our experiences. When you
meet a physician w ho seems pretty mu ch like most of the other ph ysicians you h ave met, you really do n't have to
remember m uch ab out her. 13 Theoretically, all you need to remember is that Dr. Fox is a typical doctor, and your
schema will generate inferences about her later if you need them. Obviously, if all you ever remembered was
how each doctor was like most other doctors, over time you would not be able to remember how Dr. Fox was
different from Dr. Wolf. And that sort of thing happens. What was your first-grade teacher like? You may find
that m ost of w hat yo u can remem ber (or at least infer) a bout her (an d she w as fema le wasn 't she?) is h eavily
driven by an vague set of memories you have for elementary school teachers in general. But just as clearly you
often do remember individual people you have known, and often the most salient features about them are what
sets them apart. I can still recall my first-grade teacher, and what I most remember about her was that she was
younger, mo re lenient, and more harried tha n my other elemen tary school teach ers, features that one imagin es
have some importa nt cau sal links to one anoth er.
Salience Effects: We ha ve sugg ested that in consisten t behavio rs may “ stand o ut” an d receive extra
processing that ma kes them h ighly acces sible in mem ory. Beha viors ma y also be giv en priority in m emory
process ing bec ause t hey are extreme, v ivid, or c ultura lly devia nt. Ro thbart , Fulero, Jen sen, H owa rd, and Birrell
(1978) h ave inves tigated on e clear stereotypin g conseq uence of th is greater acces sibility. If behavio rs are
remem bered by group rather th an by individ ual an d if certain group s perform behav iors wh ich are h ighly
memo rable, then it follow s that the en tire group w ill tend to be seen as likely to perform these salient b ehaviors
especially given the previously discu ssed tendency (C hapter 3) to assum e that members o f groups are similar to
one another. One way that behaviors become memorable is by being highly deviant statistically or normatively.
In their experiments Rothbart and his colleagues presented subjects with information about height or crimes by
individuals. Although the average height or crime propensity was equal for the two groups, members of one
group h ad mo re very tall people or p eople who comm itted more serio us crimes . When s ubjects w ere
subsequently asked for judgments about the groups the group with more tall members was judged as taller and
the group with members who committed more serious crimes was judged as having committed more crimes.
Thus through group identification of events the entire group tends to be ascribed the behaviors of a few deviant
individuals.
Misanthropic Memory: We have been looking at memory for what might be called raw behaviors or
features of stan dard p eople or g roups . Yet in o ur every day liv es we o ften enc oun ter inform ation a bout people
whose group memberships have affective significance for us, and often that behavior comes accompanied by
explanations of one kind or another. Oscar Ybarra (Ybarra, 1999; Ybarra & Stephan, 1996) has pointed to an
important bias in memory, namely that we tend to recall negative behaviors that have dispositional explanations
and positive behaviors that seem more situationally caused. He has termed this misanthropic memory because
the net result is that we tend to see the negative behaviors of others as a fundamental part of personality and
permanent whereas more positive behaviors are regarded as fleeting. We hold people responsible for their bad
behavior and don’t give them c redit for their positive.
Ybarra argues that this effect may occur because it helps to make people feel good about themselves by
casting o thers in a ba d light and by com parison o urselves in a more fav orable on e. In suppo rt of that idea , Ybarra
(1999) foun d that wh en people did not ha ve the motivation (because their self-esteem had been ra ised) or ability
to use this form of indirect self-enhancement, the misanthropic memory effect was erased. Furthermore such
effects also occur for memory for group behaviors. In particular misanthropic memory occurs for out-groups but
13
For a while I tended to wear light tan socks on days when I taught. Sometime later I encountered a
student whom I recognized as having been in one of my large classes. After we had done the usual
introductory thin gs, he told me that h e didn’t remember m y name or even what class h e had taken w ith
me, but he did recall that I was the guy who never wore socks to class. I would have preferred to have
been ind ividua ted in a s omew hat m ore profo und way.
Chapter 4: Schemas / Version of November 10, 2002
18
not for in-groups (Ybarra, Stephan, & Schaberg, 2000). It also follows from this perspective that when we have
positive expecta ncies a bout a perso n we m ay ass ume th at he is ju st being norm ative or at least g iven th e wellknown discounting principle from attribution theory, and therefore we might be quite uncertain about the
reasons for behaviors from generally normative people. This in turn may lead to our processing information,
particu larly un expected behav ior from such people esp ecially ca refully, an d the n et result w ould b e that pe ople
would recall nega tive behavior by pos itive people better than positive behavior by n egative people; this has been
foun d (Yba rra, Sch aberg, & Keiper, 1 999).
Inferences
Heretofore we have discussed the effects of schemata and stereotypes on attention, labeling, and
memory, but perhaps the most obvious features of schemata is that they facilitate inferences we make about
people a nd ob jects related to the sc hema (Taylo r & Cro cker, 19 81). Su rely you are quite certain th at the pe ople
you meet at a party have functioning hearts? And where is your direct evidence for this belief? Are you sure that
the library you are about to enter for the first time will have book shelves? Does the driver of the car which just
passed you have a driver's license? Is the professor you have just met liberal? Is the large man approaching you
on the dark street likely to assault you? Does your obviously gay waiter have AIDS? Obviously some of these
inferences have a stronger rationale than others. You are certain the people at the party have hearts because you
have a large store of biological and medical knowledge at your disposal; you have a theory about what humans
are like. On the other hand, the belief that a gay waiter has AIDS can be based only on probabilistic reasoning
that m ay be fla wed in any n umb er of wa ys. Still, wh en the g oing g ets rou gh or w e are no t being es pecially
reflective we infer attributes from categories whether or not these inferences are firmly justified.
Our s chem ata (an d stereo types) pro vide de fault v alues fo r missin g inform ation. W e know that cer tain
features must be or are likely to be present, and so we simply infer them even though we don't directly see them.
Sometimes these inferences are theory based and sometimes mere empirical generalization. Sometimes we make
errors. These inferences also tend to get stronger over time as direct memory traces for what we have or have not
experienced fade (Sanbonmatsu, Kardes, & Sansone, 1991). In a way the rest of this book is about such inferences,
but for the mom ent let us discuss the extent to w hich such stereotype-based inferences occu r and wh at their fate
is when they co nflict with direct behavioral eviden ce.
Stereotypes and Individuating Information
While schema theories emphasize the importance of prior expectancies and information on the
processing of new information, not even the most enthusiastic schema theorist would insist that the incoming
information itself has no role to play. The person who has a stereotype that professors are politically liberal
would still, in most cases, take note of the fact that this particular professor voted for a conservative Republican.
Similarly the pers on wh o believes tha t blacks are la zy, migh t label a wide ra nge of am biguou s behav iors
perform ed by b lack co -work ers as la zy, bu t wou ld prob ably n ot ma nage to make the wo rk of Jon es, a no toriou sly
hard work er, fit her stereo type.
The Importance of Behavior
Still, while everyone would, in principle, agree that our perceptions, memories, and inferences are joint
products of ou r pre-existing knowledge and incoming da ta, in practice most stereotype researchers h ave focused
their resea rch atte ntion mostly on the schem a end of the bu siness. T his view was c halleng ed quite strikingly in
the early 1980 s in a ser ies of pap ers by A nne L ocksley and h er colleag ues.
Locksley's Challenge
The initial set of experiments by Locksley, Borgida, Brekke, and Hepburn (1980) seemed to suggest that
stereotypes had a role to play in inferences about others only when there was no other behavioral information
available. In an initial study subjects were told that a male, a female, or someone not identified by gender had
performed either assertive or passive behaviors across three different situations. On ratings of assertiveness, the
gende r label ha d no effe ct wh ereas th e actu al beha vior did . That is , subjects were no more like ly to see a male
than a female as assertive, but were willing to say that a person who had been assertive was more assertive than
a person who had been passive. In a second study, the diagnosticity of the behavioral information was varied;
some of the behaviors strongly implied assertiveness whether other behaviors did not. Again, when the
behavioral inform ation was d iagnostic for the rating (as it wa s in the first study), gender label played no role (See
Table 4.1). How ever, subjects did infer that the male stimu lus person wa s more assertive than the female when
the behavioral information could not be used to judge assertiveness or no behavioral information was presented.
Chapter 4: Schemas / Version of November 10, 2002
19
They are also inclined to use gender when behavioral information is mixed so that the target performs equal
numbers of assertive and passive behaviors (Locksley, Hepburn, & Ortiz, 1982a) . These studies seem to suggest
that subjects prefer to infer traits from behaviors and will use gender only as a last resort, only if they have no
good behav ioral info rmatio n to go on.
Tab le 4.1 : Attributions of Assertiveness (Locksley, et al., 1980, Experiment 2)
Behavioral Information
Sex of Target
Male
Fema le
Gend er Only
49.44
46.59
Gend er + No n-Diagn ostic
48.73
44.05
Gend er + Dia gnostic
67.62
67.30
On the face of it this is a striking set of results tha t has broad im plications for the use of stereotypes. Th ey
would , for example, su ggest that s tereotypes play a role in ou r perceptions o f others on ly in those fa irly rare
cases where we have no additional information about them. So a racist may think that Hispanics are lazy but be
willing to give up that belief at the first hint of hard-w ork by Jose. At their most d amaging th en stereotypes
would be nothin g worse th an initial, “so ft” hypoth eses that ca n easily be ov ercome by "hard" d ata. Inde ed, there
are no w sev eral stu dies us ing a v ariety of re search metho ds sh owin g that s tereotyp ic inferenc es are sev erely
reduced or eliminated w hen additiona l individuating inform ation is also present (e.g., Del Boca & A shmore,
1980; Dipboye & Wiley, 1978; Eagly & Steffen, 1984; Glick, Zion, & Nelson, 1988; Jussim, Nelson, Manis, & Soffin,
1995; M acrae, Shep herd, & M ilne, 1992; M ettrick & Co wan, 1 996; W eisz & Jon es, 1993).
Is Locksley Correct?
Obviously there are important theoretical issues underfoot here, and there are also important legal
implication s for how we think a bout dis crimination (Borgida, R udma n, & Ma nteufel, 199 5). If stereotypes a re
such soft and fragile things, why subject them to the proding of the legal system? A lot depends on what
Locksley and her colleagues wa nted to claim for these da ta. If all they meant to say w as that som etimes
individu ating da ta such as beha vioral or trait info rmation takes preced ence over exp ectations, o ne wou ld be hard
pressed to find anyo ne who w ould disagree. If, on the oth er hand, they w ere suggesting that indiv iduating data
always or even usually take precedence over schema information and stereotypes, then they were plainly wrong
and wrong in ways that do not require elaborate experimental demonstrations. If our schemata, our expectations,
our stereotypes are merely cogn itive feathers that get blown aw ay by the wind s of empirical data, it is hard to
imagine that they would have been taken as seriously for so long by so many social scientists and the people who
have been victimized by them. The very concept of stereotype was invented to label a common element of our
experience, namely that ou r expectations, attitudes, prejudices, stereotypes, and th eories bias us and do so even
in the presence of valid disconfirming behavioral data. We are not always fully responsive to behavioral data.
This is not a m atter for expert pronouncem ent but of com mon experience.
Sometimes Stereotypes Matter: And lest there be those who are not convinced by argument, let us be
clear that there are abundant data supporting the idea that category (stereotype) information often does have
some weight in decisions we make about others even in the presence of individuating, behavioral data. For
examp le, studies on gen der (e.g., De aux & Lewis, 1 984; D ipboye, F romk in, & W iback, 1 975; D ipboye, A rvey, &
Terpstra, 1977; Jackson & Cash, 1985; Swim, 1993) show that gender remains a predictor of evaluations and job
suitability ratings even when there is information present about qualifications of the individuals. Labels and
behavioral or trait data both contribute to evaluations of targets with ethnicity and race (Bodenhausen, 1988;
Bodenhausen & Wyer, 1985; Gordon, 1990; Gordon, Bindrim, McNicholas, & Waldren, 1988; Macrae & Shepherd,
1989a, 1989b), homosexuality (Laner & Laner, 1979; Millham & Weinberger, 1977; Storms, 1978), mental
retardation (Skowronski, Carlston, & Isham, 1993), age (Vrugt & Schabracq, 1996), physical attractiveness
(Budesheim & DePaola, 1994), criminality and alcoholism (Hannah & Midlarsky, 1987) and mental illness (Link
& Cullen, 1983) . So -- sometimes individual attributes are more important than stereotypes, sometimes it is the
reverse, a nd so metim es they in teract in all man ner of w ays. N ot, perha ps, a big surprise .
One might be tempted to perform a meta-analysis of such studies to see whether stereotypes or behavior
are, on average, more important. However, it is hard to offer strong generalizations about the relative power of
Chapter 4: Schemas / Version of November 10, 2002
20
group designation and behavioral qualifications because so much depends on how powerful or salient each
factor is. Few companies would want to hire totally unqualified males over highly qualified females, so that
qualifications probably are determinative when strong. Even the most sexist person is bound to recognize the
outstanding achievements of his female colleagues although, of course, he may minimize their importance or find
other ways to protect his general sexist beliefs. On the other hand, there are bound to be situations where race
and gender are powerful factors. One would be the case where the perceiver has strong and overt biases.
Generally, I imagine, it takes stronger evidence to convince a sexist, with his strong stereotype, than a non-sexist
that females really can do complex mathematics well. Another might be in those cases where gender might be
cons idered, by some , to be an importa nt qua lification for a job . One im agines that m ost clo thing s tores w ould
have strong preferences for h iring women o ver men to staff their lingerie department, and in jobs requiring brute
strength those who hire and fire may let their gender stereotypes about such things dictate their decisions. As I
have said and will continue to say, questions posed in terms of which factor is the more powerful or important
are alm ost alw ays u nans werab le in wa ys tha t mak e sense. I t all depen ds.
Another reason it is hard to say whether stereotypes or behaviors are more important is that they may
differentia lly affect d ifferent so rts of infere nces o r decisio ns. Ku nda a nd Th agard (1996 ) argue t hat for trait
ratings behav ioral info rmatio n tend s to do mina te stereoty pes wh en the b ehav iors are a t least m inima lly
informative because the link between traits and behaviors is stronger than the link between stereotypes and
traits. However, for more complex judgments such as job suitability, predictions of future behaviors, or
attributional judgments stereotypes often dominate because of close ties between the fairly abstract stereotype
and these more abstract types of judgments which are less likely not to be affected by particular behaviors.
When Stereotypes are Used: The issue is, of course, and should have seen as such from the outset, not
which kind of information is mo re important but w hat kinds of circum stances favor o ne type of information o ver
the other. Before discussing that, however, let us consider some possible relationships between stereotypes and
behavior. Bodenhausen and Wyer (1985) have suggested that there are basically three ways that stereotypes and
individ uatin g inform ation m ay be rela ted, an d I will a dd sev eral mo re.
The Default Perspective: The default perspective is close to that of Locksley and her colleagues.
Stereotypes act as assumptions we make about others in the absence of more useful information, but they will be
abandoned at the first hint that the person violates those assumptions. As we will see shortly this model may
work best for tra it judgm ents (see p . ?), but in its genera l form th is positio n seem s unte nable.
The Heuristic Perspective: The heuristic position suggests that stereotypes act like any means of
simplifying a complex w orld -- they are shortcut rules, easy to use, that owe their existence to the fac t that they
work at least part of the time. These heuristic rules will be used until unambiguously inconsistent information
shows up. In other words rather than being dropped at the first hint of incoming data, heuristic stereotypes must
be actively disconfirmed. Th is is close to the traditional view of stereotypes seen as abstract kno wledge structures
that a ctively res ist cha nge.
Stereotypes and B ehavior as Conceptu ally Equivalent: There are at least three other possible ways that
traits and stereotypes could be related. Kunda and Thagard (1996), among others, completely abolish the
category-feature distinction. In this view the stereotype does not have a privileged status in controlling the
processing of informa tion; behavioral an d category informa tion have the sa me status. Thu s, knowing th at a
person is a colleg e profess or ma y or ma y not b e usefu l but ha s no in herent a dvan tage ov er know ing tha t she is
smart, tall, and drives a Mazda. Not only does this provide for a certain simplicity of argument and economy of
explanation for results, but it also surgically removes an important but under-explored problem about the
relationships between features and categories (see Chapter 3). To be sure, it seems natural to think of features as
associated with culturally or biologically salient categories. So, dogs have tails, professors are well-educated, and
cab drivers are rude. But there is no principled reason why we could not reverse this. Dogness is a feature of
animals with tails, well-educated people (category) are often professorial (feature), and rude people include cabdrivers.
While I am sym pathetic with the Ku nda-Tha gard argum ent about ab olishing the category-attribute
distinction (and h ave stressed this in Ch apters 1 and 3 ), I also think it important to distingu ish between catego ries
as types of information and categories as schemata which may affect attention, labeling, memory and inferences.
For cognitive or cultu ral reasons som e categories have a rich featu re set and a long cu ltural history of being used
as the b asis of s tereotyp es. Peop le define th emselv es as m embers of socia l catego ries but ty pically n ot of tra it
catego ries, and social c ategor ies are typ ically us ed to d istingu ish betw een gro ups, w hereas feature o r trait
categories are typically used to distinguish individuals (Bodenhausen, Macrae, & Sherman, 1999). Furthermore as
Chapter 4: Schemas / Version of November 10, 2002
21
I discussed in the last chapter it is often awkward to see socially defined categories as nested in a hierarchical
fashion. Others (e.g., Brewer & Harasty Feinberg, 1999) have argued for a principled distinction along these and
other lines. Having said all that, however, it is not inevitable that this distinction is always important, and in the
case we are discussing now – drawing inferences about individuals – it may make more sense to see the
distinction as mo re implicit than active.
Stereotypes and Behaviors Interact: Another possibility is that stereotypes change the meanings of
behav iors. O ne of th e tangles we are a bout to enter (s ee p. ?)is tha t behav iors them selves m ay ch ange th eir
meanings or be evaluated differently as a function of whether a stereotype is salient. Just as roles and situations
affect th e way s we co nstru e behav iors so c an stere otypes .
Behaviors May Trigger Stereotypes: Just as stereotypes may a ffect the meaning of beha viors, sometimes
behaviors can affect stereotypes. It is easy to assume that stereotypes sit waiting like trolls to capture and molest
unsuspecting behaviors. After all stereotypes are a fairly stable part of the mental apparatus we carry around
with us, and behaviors happen before our eyes. Yet, as we indicated in our earlier discussion of priming,
stereotypes are neither stable nor always available. If they’re trolls, they sometimes take a vacation. I talk to my
professional black neighbor. During our conversation his race is not especially salient to me and never once do I
remotely enterta in the thou ght that h e is stupid or v iolent or wh atever black stereotypes a re in the air.
One wa y in which this h appens is that beha viors may key o nly parts of a given stereotype or stim ulate
particular sub-types. It is a part of Sarah’s stereotype that Mexican-Americans are violent and family-oriented.
Whe n she o bserves a large ex tended Mexica n-Am erican fa mily ha ving a holida y picnic at the pa rk, it is sure ly
more likely that the family-oriented part of the stereotypes will be more salient that the violent part, and the
latter may even be inhibited. Alternatively, w e might argue tha t Sarah has stereotypes of two distinct su b-types
of Mexican-Americans and that family behavior makes one salient and watching a young Mexican-American boy
swagger abo ut makes the o ther more salient.
Many of our stereotypes are not salient for particular people, and sometimes they are held in abeyance
while we check this person out. In that case, it is possible that one (or a very few) stereotypic behaviors may
themselve s make th e stereotype sa lient. I’ve learned never to as sume th at my pro fessorial colleag ues are
politically liberal or no t religious or th e owners of old cars . Some are, bu t there are lots of exc eptions. H owever,
these stereotypes are loaded and ready to fire should an appropriate trigger be displayed. So if a new colleague
anno unce s he is su pports a liberal politic al can didate , I may be all too ready to assu me tha t he also drives a n old
car, shops at a health store, does not regularly attend church, and has a large library at home. To be sure many of
these features are thems elves related (liberals m ay well drive o lder, or at least ch eaper, cars tha n cons ervatives).
But I have argued (Schneider & Blankm eyer, 1983) that certain stereotypes may produce even closer relationships
among relevant features (see Chapter 5 for furth er discussion), so tha t a triggering behavior ma y not only m ake a
given s tereotyp e more s alient bu t lead m ore read ily to stereo typic infer ences.
When Are Stereotype s Potent?
Stereotype Strength: Stereotypes are not any one thing, always salient, never sleeping, always alert and
potent. They ebb and flow. So a given stereotype will be relatively strong under some circumstances and at some
times but not oth ers. In Chapter 1 I su ggested that little attention has been g iven to measu ring the actual strength
of stereotypes, certainly a relevant concern when we want to compare the stereotypes a given person has of
various groups. But that lacunae in ou r understanding haun ts us here as well. In thinking about stereotype
strength it is sometimes impo rtant to distinguish b etween stereotype activation (whether the stereotype is salient)
and whether it is applied to an individual or group (Macrae & Bodenhausen, 2000). It is certainly possible for the
first to occur withou t the second, althou gh I suspect tha t in everyday life the two are so closely linked that they
may be insepara ble.
Relative Strength and Stereotype Functions: Schemata put their best foot forward when they help us
und erstan d am biguo us info rmatio n, and so as p reviou sly disc ussed (see p. ?) stere otypes will likely be fa vored if
behavioral or individuating information is ambiguous. When a person needs to make a quick decision or has a
“lot on her mind” stereotypes should be especially pertinent. Several studies (e.g., Gordon & Anderson, 1995;
Hadjimarco u & Hu , 1999; Macra e, Hewstone, & G riffiths, 1993; Martell, 1991; Pend ry, 1998; Pend ry & Macrae,
1999; Pratto & Bargh, 1991; van Knippenberg, Dijksterhuis, & Vermeulen, 1999; White & Carlston, 1983) have
shown that when subjects are cognitively busy or have to make quick decisions, they tend to reply more on initial
expectations and stereotypes than on actual behav ior in making their judgm ents. On the oth er hand wh en
cognitive resources are available and people are motivated to form accurate impressions because of uncertainty,
stereoty pe use is dimin ished (W eary, Jac obso n, Edw ards, & Tobin , 2001 ).
Chapter 4: Schemas / Version of November 10, 2002
22
Bodenhausen (1990) argues that stereotypes should be used more when people are not at peak mental
efficiency and thereby inclined to use co gnitive sho rtcuts. Th is was fou nd to be tru e when “ night” peo ple were
asked to judge race-stereotypic crimes during the day or when “day” people did the task at night. People use
stereotypes m ore when they hav e been deprived of contro l on a previo us task (R ic, 1997). A lso when people are
aroused they tend to rely more on stereotypes (Paulhus, Martin, & Murphy, 1992), an example of the well-known
fact that arousal tends to restrict attention and encourage reliance on salient cues. The net result then is that
stereoty pes are pr obab ly more importa nt tha n beha vioral d ata w hen peo ple are rus hed, bu sy, ha ssled, o r simply
lazy. One interesting an d somew hat parado xical implication is that the use of stereotypes might be encou raged
by having a rich array of individuating information because so much information leads to cognitive overload
(Brewer, 199 6).
Category Fit: Self Categorization Theory (Chapter 3), dual processing theory (see p. ?) and the
continuu m mod el (see p. ?) all argue that, at least under some circu mstances, w hen people seem to fit a
stereoty pe categ ory w ell, they w ill be so ca tegorize d and relevan t stereoty pes read ily applied . It is certain ly
pertinent to raise the question of ho w close the ft is before such n on-reflective categorization takes place, bu t it a
reasonable hypothesis that people who look like or act like prototypic members of categories will fall victim of
stereoty pes ass ociated with th ose ca tegories.
Motivational and Emotional Factors: While stereotype research has tended to downplay the importance
of mo tivation , obvio usly th ere will be tim es wh en we w ant to use or s tereotyp es and other tim es wh en we w ill
want to inh ibit our stereotypes. Ziva Ku nda (e.g., Klein & Ku nda, 1992 ; Kunda & Sinclair, 1999) argu es that often
the ways we process information about others is affected by our desires to develop and maintain a particular
impress ion of a target. In such cases p eople w ill deploy th eir stereoty pes or in hibit the m in w ays th at help
support the desired impression. The ex-husband’s new wife is good looking and therefore must be unintelligent
and superficial. The daughter of the old friend is wearing multiple noserings and you have to inhibit your
largely negative stereotype abou t people who wo uld do that.
In one demonstration Sinclair and Kunda (2000) found that although students who received a high grade
evaluated fema le slightly more positively than ma le professors, when they h ad received a low g rade they
evaluated the female professor more negatively. A quite reasonable interpretation of these results is that when a
person’s eg o is on th e line, they ma y use gen der (in this ca se) stereotypes to excuse the o utcom es they get. Th ere
are actu ally seve ral dem onstra tions o f recruitin g stereot ypes to serve m otivat ional n eeds (See Cha pter 8).
It is com mon ly assu med a nd am ply dem onstra ted tha t our st ereotype s of ou t-grou ps are es pecially
strong when there is threat from the out-group (see Chapters 7 and 8). As one example, although activation of
black stereotypes for whites tends to dissipate over time, it tends to be maintained when there is a disagreement
between a white perso n and a black (K unda , Davies, Ad ams, & S pencer, 200 2). Stereotypes o f the enemy are
particular strong during wars and economic conflicts. A related idea is that some people adopt strong
stereotypes of some groups as a kind of defense mechanism; this was the idea behind the Authoritarian
Perso nality (s ee Cha pters 1 a nd 8).
Over and above ego-involvement, appearance or obvious group membership cues will give rise to strong
feelings of initial like or dislike in many cases. One possibility is that it is easier to form and use stereotypes of
groups with which have traits that depart from normal expectations and which are perceived to have high
hom ogene ity (Dijkste rhuis & van K nipperb erg, 199 9). For m ost of u s seeing a skinh ead w ith a sw astika on his
shirt will not only give rise to immediate feelings of loathing but will “give us permission” to activate our
stereotypes of such people, to assume they have a wide range of stereotypic features associated with the category,
and to interpret his behav ior in stereotype consistent wa ys. By contrast there are circu mstances w hen we wa nt to
be sure o ur judg ments are no t based on ou r stereoty pes. In s ome c ases th ese mo tives stem from d eeply held
person al beliefs a bout equality and th e import ance o f individ uals (s ee Cha pter 8 for fu rther dis cuss ion), bu t in
other c ases w e may be aw are of n orma tive press ures fro m oth ers that can a ffect wh at inform ation w e deploy in
our judgments. The social judgeability model of Yzerbyt and his colleagues (Yzerbyt, Leyens, & Schadron, 1997)
predicts that we use information about people in judging them only when we think such use is cognitively,
morally, or normatively justified, and clearly there are many circumstances when we think that stereotype
inform ation is not.
Importance of Behavior
Obviously behavioral information can seem relatively strong or weak as well. It is likely that seeing a
priest bea ting a c at to de ath w ould o verwh elm wh atever p ositive s tereotyp es you have o f men o f the cloth in
terms of kindness. Extreme behaviors stand out and are probably weighted especially highly in judgments about
Chapter 4: Schemas / Version of November 10, 2002
23
others. In Chapter 3 I discussed the important work of Monica Biernat and her colleagues on shifting standards
(cf, Biernat, Vescio , & Man is, 1998). Th e basic argu ment is tha t a behav ior or feature w hich wo uld be no rmal for a
person from o ne cate gory m ay seem quite extre me for s omeo ne in an other c ategor y. A perso n wh o is 6 feet ta ll
would be seen as more or less avera ge if a male but quite tall if a female. In one instructive set o f studies
(Kobrynow icz & Biernat, 1997 ) relevant to preset concerns su bjects indicated that fathers an d mothers perform ed
the same general types of behaviors. Fathers and mothers who described their parental activities as good or
“alright” parenting were rated as being equivalently good, but when asked how many behaviors each parent
performed they estimated that the good mother performed more than the good father. Thus it takes fewer good
behaviors to be a g ood father than a good m other. In judging ho w importan t behaviors are we h ave to take into
accoun t the kind of person w ho performs them as well as how extreme they are.
One way to approach this is to think about how informative or diagnostic various kinds of information
are likely to be. Rasinski, Crocker, and Hastie (1985) have argued that Locksley and her colleagues assumed that
assertive beh aviors w ould be eq ually diag nostic for m ales and females Is th is true? Sup pose tha t females are
inhibited from being assertive, and males are encouraged to be so. Now according to standard attributional
principles of inform ation g ain (Jon es & M cGillis, 19 76), if we observ e an as sertive be havio r perform ed by a male
we have n ot learned very mu ch abou t his unique assertivenes s -- he is about like all males, just wha t we expected
on th e basis o f our ste reotype. B y con trast th e assertiv e female is perceived to be ver y asser tive ind eed; after a ll
look at all the cultural conditioning she had to overcome to perform that behavior. As Biernat, Manis, and Nelson
(1991 ) have a rgued , the sam e assertiv e behav ior ma y be seen as mo re extreme when it is com pared to a fema le
rather th an to a male sta ndar d.
Think about perceptions of assertiveness as an equation with gender and behavioral information
contributing according to their weights. Maleness would lead the subject to believe the person would be assertive
(giving gender a high weight) whereas the assertive behavior by the male adds relatively little to the final
judgment (low weight for behavior). For females, however, gender information predicts low assertiveness (low
weigh t), but th is is com pensa ted for by the pow erful beh aviora l inform ation w ith its hig h weig ht. Lo a nd beh old
subjects see the two as equally assertive, but this is hardly a surprise. We have two pieces of information about
two pe ople, an d in eac h cas e one piec e stron gly lead s us to believe th at the pe rson is assertiv e. The larg er point is
that co ntrary to the m essage of Loc ksley an d her co lleagues , gender is hard ly irrelevan t in this a nalys is. It is
relevant in the first instance because it leads us to make direct inferences for the male, and it is relevant in the
secon d insta nce bec ause it lea ds us to interpr et the fem ale's beh avior a s especia lly assertiv e.
Relative Diagnosticity: The more general po int has been ma de by Krueger an d Rothba rt (1988). They
argue the reasonable case that category information and behavioral information may have independent
diagnostic va lues and these m ay vary from situ ation to situation, perceiver to perceiver, judgm ent to judgmen t.
In the Locksley case gender may not have been highly diagnostic because people do not expect that males and
females differ all that much in a ssertiveness. By contras t the behavioral inform ation may h ave been seen as qu ite
diagnostic, especially for females. Thus, Locksley and her colleagues compared a case where the behavioral
inform ation is more d iagno stic tha n the ca tegory in forma tion. H owev er, one co uld ea sily ima gine situ ations in
which the opposite were true. For example, one would generally be better off using stereotypes based on body
type to judge whether someone might make a good football lineman than in whether he can effectively growl as
such folks are prone to do during games.
Krueg er and Roth bart u sed info rmatio n for w hich c ategor y was demo nstrat ively dia gnos tic, nam ely
aggressiveness14, but they also manipulated category diagnosticity. Subjects were told that a male or female had
perform ed the b ehav ior or th at a ho usew ife or con structio n wo rker had , on the assu mption that su bjects w ould
assume that there would be greater differences in aggressive behavior between housewife and construction
worker than between male or female so that occupations would be even more diagnostic of aggression than
gender. They also used behavioral information that had been rated to be highly or less highly diagnostic of
aggres sivenes s. The res ults sh owed that bo th sou rces of in forma tion ha d effects in propor tion to their
diagno sticity. That is d iagnostic behavio ral informa tion affected ra tings of ag gressivenes s and s o did the c ategory
informatio n especially w hen con trasting ho usewife vers us con struction worker.
14
Certainly actual agg ressive behavior as o pposed to that w hich is merely assertive (as the in Lo cksley
studies) is m ore highly c orrelated w ith gender.
Chapter 4: Schemas / Version of November 10, 2002
24
In subsequ ent experiments Krueger an d Rothba rt (1988) show ed that the category effect cou ld be washed
out when the behavior was consistent in aggressiveness over situations and hence indicated an aggressive
disposition. That is, when subjects were told that a person was consistently aggressive they tended to let that
information rather than gender affect their judgments. Thus, a single bit of behavior will not necessarily have
precedence over category information but will if and when behavior gives evidence of a strong disposition.
Sensible enough. Suppose you are predicting whether I can hit a free throw in basketball, and you have two
sources of information. You know that I am or am not a member of a major basketball team, and you know
wheth er I hav e just hit a previou s free thro w. No w give n tha t I migh t have been lu cky in m y previo us atte mpt it
might well be quite rational for you to make you r prediction on the basis o f whether I am a regular player or not.
However, if you now are told that although I am not a regular player I just hit the last 10 free throws I attempted,
you would probably be wise to pay attention to my past behavior. Note that this does not represent any sort of
proposition abou t either sort of information being m ore important than the other. Rather it recognizes tha t both
have their value as predictors and tha t a rational person w ould pay atten tion to which is th e more important.
While we hav e focused on perceived diagnosticity of informa tion, salience plays an an alogous role.
Intuitively, a vivid portrayal of one person beating up another (say via video tape with lots of blood and
appropriate sounds) is going to be given more weight than a brief description such as “beat up someone in a
parking lot”.15 Similarly, category membership might be made more salient through appearance cues or frequent
reminders. Beckett and Park (1995) argued that gender may not have been especially salient for subjects in the
original Locksley research sin ce gender information was con veyed verbally rather than pictorially. So they
conc eptually replicated the stu dies by g iving su bjects info rmatio n abo ut a m ale or fem ale wh o wa s different ially
assertive across several situa tions. With a m ere verbal description of the target as ma le or female, gender played
little role in predictions o f assertivenes s in future situ ations; this replicated the L ocksley resea rch. How ever,
when subject s also s aw a picture o f the perso n (thu s presu mably makin g gend er more s alient), su bjects did
predict that the male would be more assertive than the female although gender was still less important as a
predicto r than actua l behav ior.
Finally the ambiguity or ease of using information may play a role. For example, Nelson, Biernat, and
Manis (1990) asked subjects to judge the heights of men and women. Not surprisingly men were perceived to be
taller than women even when actual height was statistically controlled. However, gender stereotypes about
height played a larger role wh en the men an d wom en were seated than when stan ding, because ap pearance data
are mo re amb iguou s in tha t case.
The Role of Non-Diagnostic Information: Often the diagnosticity of category or behavioral information
is readily appare nt, but some times peo ple may misjud ge these matters . People w ho driv e old ca rs are pro perly
outra ged w hen th ey are sto pped in “nice” neighb orhoo ds by p olice wh o seem to feel tha t old ca rs are dia gnos tic
of bad intentions and a thieving disposition.16 Man y people a re unn ecessa rily frighten ed of peo ple who talk
aloud to themselves in public bec ause they assum e such peo ple may be d angerou s and th ey give the beh avior a
higher diagnostic value than they should.
Several experiments have shown that people sometimes use non-diagnostic information in a way that
reduces the effects of category inform ation (DeDreu, Y zerbyt, & Leyens, 199 5; Denhaerinck, L eyens, & Yzerbyt,
1989; Nisbett, Zukier, and Lemley, 1981). In a series of experiments Hilton and Fein (1989) examined the role of
pseudo-diagnostic information in such tasks. Pseudo-diagnostic information is information that may generally be
relevant for a wide variety of judgments which is not for the judgment at hand. For example, intelligence or GPA
may be a useful bit of information to know about someone for many tasks, but it probably does not predict how
15
Recently Texas executed a female murderer who had given clear and convincing evidence of
repentance and a dramatically changed disposition while on death row. While there were many moral
and legal issues in the case, it was striking that the defenders of her execution continually mentioned the
fact that sh e had bru tally murd ered two peo ple by repeatedly s triking them with a pick ax. One w onders
whether their reactions m ight have been so ftened had she m urdered in a wa y that is less easy to
visualize.
16
That happened to once. An initial sense of outrage quickly was rather quickly replaced by some
degree of amusement, but then since my middle class bone fides are clearly in order I have the luxury of
not w orrying abou t it.
Chapter 4: Schemas / Version of November 10, 2002
25
helpful someone would be in most situations. In one of their studies subjects were asked to judge the
assertiveness of males and females with clearly diagnostic behavioral information (telling a seedy character to go
awa y), clearly u ndiag nostic (and irre levant ) inform ation (fin ding m oney in a pho ne boo th) or ps eudo -diagn ostic
information (having a high GPA). Subjects saw males as being more assertive than females when the behavioral
information was clearly irrelevant but when diagnostic or pseudo-diagnostic information was used, males and
females were seen as equally assertive. Additional work by Fein and Hilton (1992) suggests that tendencies of
people to vote fo r cand idates o f their ow n politica l party ca n be red uced b y the pres entatio n of pse udo -diagn ostic
information (e.g., stays in touc h with old friends) in part beca use such info rmation increases th e perceived
prototy picality o f the opp osing cand idate.
One reason that pseudo-diagnostic information may dilute stereotype-based inferences is that it reminds
people th at they are judg ing an individ ual an d that they sh ould n ot use stereoty pe inform ation a s a part of their
judgm ent. Acc ording to the so cial judg eability m odel of Y zerbyt a nd his colleagu es (Yze rbyt, Ley ens, &
Schadron, 1997) such norms play a large role in the hows and whens of our use of information. Irrelevant
inform ation d oes no t trigger su ch no rms bec ause it d oes no t sufficien tly indiv iduat e the perso n.
Such results do not dispute the general point we have been making. People are generally not dopes.
Sometimes they u se stereotypes based on category mem bership to make jud gments abo ut people, and som etimes
they rely m ore hea vily on behav ioral an d oth er individ uatin g inform ation. T he relativ e import ance o f each is
highly dependent on perceived diagnosticity. That does not mean that people are always accurate in judging
diagnosticity or that they invariably perform the calculations correctly.
Stereotypes Affect Diagnosticity and Interpretations of Behavior: Up to now w e have cons idered
behaviors and stereotypes as more or less independent contributors to impressions of people. However, the two
may also interact. Suppose you read, as did the Krueger-Rothbart subjects, that a housewife or a construction
wor ker hit som eone . Wh at kin ds of men tal im ages do yo u form ? Perh aps w hen y ou th ink a bou t the h ous ewife's
hitting you imagine her spanking a naughty child, but when you imagine the construction worker hitting
someone images of fists and blood come to mind. We certainly do imagine different types of people exhibiting
traits and achieving goals in d ifferent ways (Kund a, Sinclair, & G riffin, 1997; Slu sher and Anderso n, 1987 ).
Not only do we imagine different behaviors for the same trait depending on the actor, the same behavior
can lead to different trait inferences for different actors. Dunning and Sherman (1997) used a memory paradigm
to investigate tacit inferences that people make when they read about behavior. For example, if you are told that
the head of a computer company needed to terminate a few employees, you would probably assume that
metaph orical but n ot actua l heads w ould roll as people were fired. H owever, if yo u were told that the h ead of a
drug cartel decided to terminate a few employees, you would more likely red blood than pink slips. Dunning
and Sh erman pres ented su bjects with a mbiguo us senten ces and then tested rec ognition memo ry. Subjects w ere
quite good at recognizing the sentences they had actually read, but they also recognized more sentences they had
not read th at were con sistent with the stereotypes . Given the exa mple abo ve, subjects w ho read th at the hea d of a
computer com pany terminated employees wou ld more likely think he had read that the head of the compu ter
company fired employees rather than a sentence stating that he had killed them. Those who had read the head of
a drug cartel had terminated employees w ould be mo re likely to recognize (falsely) a sentence that he killed
rather than one where he fired employees.
To return to the example of the construction worker and housewife, one possible reason subjects thought
that the construction worker who hit was more aggressive than the housewife was because they imagined the
construction worker actually behaving more aggressively. Kunda and Sherman-Williams (1993) showed that
stereoty pes affect ed judg ments when such behav ioral info rmatio n wa s amb iguou s (as in th e exam ple just giv en),
but when the behavior was specified more directly so subjects could not imagine different kinds of aggressive
behaviors, gender no longer played a role in judgments.
Stereotypes and diag nostic information may also interact in ev aluative judgm ents. Category-based
expectancy theory (Jussim, Coleman & Lerch, 1987; Jackson, Sullivan, & Hodge, 1993) proposes that behavior
which violates group-based expectancies will be evaluated extremely in the direction of the behavior. If we have
high expectations we are especially negative when the person does poorly and if our expectations are negative
we especially value positive performance. So if we expect Asians to be polite we will evaluate a cussing,
swearing Asian especially negatively (more negatively than a white person performing the same behavior) and
we wou ld evaluate an a ward win ning black scientist mo re favorably than h er equally competent w hite
colleagues. Well behaved African-Americans are a “credit to their race” and brutish whites are just “white trash”.
There is n ow co nsidera ble eviden ce in fav or of th is notio n (Betten court , Dill, Greath ouse, C harlto n, &
Chapter 4: Schemas / Version of November 10, 2002
26
Mulholland, 1997; Coleman, Jussim, & Kelly, 1995; Jackson, Sullivan, & Hodge, 1993; Jussim, Fleming, Coleman,
& Koh berger, 1996 ; Kernah an, Barth olow, & Bettencou rt, 2000). O ther mod els which we will discu ss in Ch apter 7
argue that w e should polarize eva luations of ou t-group (Linville & Jones, 198 0) or in-group (Marqu es & Yzerbyt,
1988 ) memb ers, but ex pectan cy viola tion effect s occu r for judg ments of both in-grou p and out-gro up peo ple
(Kernahan, Bartholow, & Bettencourt, 2000). There are important practical implications of this idea in admission,
hiring, and prom otion decisions, bu t we bring up expectan cy-violation theory in this co ntext because it is anoth er
illustration o f the propos ition that be haviors chang e their mean ing depend ing on w ho perform s them; beh aviors
and s tereotyp es intera ct.
Processing Concerns
In general as it has become clear that there is no general answer to the question of whether schemata and
stereotypes a re more or less im portant th an beha vioral inform ation, atten tion has shifted to the ways b oth are
used in pro cessing info rmation about p eople. Some a uthors h ave focu sed on c ognitive fac tors wh ereas others
have been m ore interested in motivationa l effects. There have been three prominen t models that ha ve attempted
to explain how our stereotypes and individuating information contribute to our impressions of individuals.
Sequential Processing Approaches
Marilyn Brewer (1988) and Susan Fiske and her colleagues (Fiske & Pavaleck, 1986; Fiske & Neuberg,
1990 ) have p ropos ed simila r mod els of ho w we p rocess inform ation a bout others . Both m odels s ugges t that in
forming impressions of others we go thro ugh a defined, yet co mplexly inter-related, set of processes. Despite
many common elements, however, the models also differ in important ways.
Brewer’s Dual Process Model: Marilyn Brewer’s du al process mod el (Brewer, 1988; Brewer & H arasty
Feinstein, 1999) is based o n the idea that there are two different routes to forming impression s. In what is called
category-based processing we respond to people primarily in terms of whatever categories we place them in,
wherea s in perso naliza tion pro cess w e are mo re respon sive to in dividu al featu res.
Categorization: In an initial, autom atic, stage, the person is identified by assessing a ge, gender, and race.
This initial identification does not lead to stereotypes. In Chap ter 3 we reviewed evidence th at supports the idea
that such ca tegorization is frequently, perhaps alw ays, autom atic at least for some catego ries. At this initial stage,
the perce iver also catego rizes the person as releva nt or irrelev ant to present n eeds. Th e vast m ajority o f people
whom we enco unter mo st days a re not, of cou rse, particularly relev ant, and in that ca se we pay them no more
heed a nd so not pro cess ad ditiona l inform ation a bout them (R odin, 1 987).
If the person is relevant or at least captures our attention, we may then try to place her in additional
categories. Brewer suggests th at we often use su b-types of larger categories or compo und catego ries (see Chapter
3). So ra ther tha n class ifying th e person as an Asian and th en furth er as a s tuden t, I will imm ediately p lace her in
the com pound category o f Asian stu dent. Brew er also specu lates that m ost of ou r most sa lient categories a re
represented mentally in terms of ima ges rather than sem antic propositions. Th us I will make an immediate
identification of this woman before me as an Asian student because she resembles other Asians, and it is not
necessary for me to do any elaborate processing in terms of identification rules. In Chapter 3 I discussed various
ways in wh ich som e catego ries beco me mo re usefu l in a give n situa tion th an oth ers, and those ru les apply here.
Individuation : If the target person seems to fit the category well, and her features seem to match the
stereotype associated w ith that category, then sh e will be seen in those stereotyped wa ys. If, however, she does
Chapter 4: Schemas / Version of November 10, 2002
27
not match an available category, processing will continue through various sub-types until a match is found, and
if this do es not w ork oth er majo r catego ry families will be tried . Some times, h owev er, while th e person is
success fully catego rized, behav ior or other a ttributes do not ma tch the featu res assoc iated with th at catego ry
well. In that case, the target may be individuated, by which Brewer means that she is seen as a member of the
category but as one whose features do not match well – the target will be seen as a special case. So as the
perceiver c ontin ues to try to find an ap propriat e catego ry for the myster ious ta rget, even tually th e target w ill
either been seen as a good exemplar of a general category, an exemplar of a sub-category, or as a member of
some catego ry with individ ual an d som ewha t peculia r features .
Personalization : There is a n altern ative ro ute to fo rming impress ions, h owev er, Whe n the per ceiver is
highly motiv ated to “get to k now ” ano ther, sh e may engag e in wh at Brew er calls pers onaliz ed proc essing. T his
Figure 4.2 : Brewer’s models of Individuation and Personalization
does not occur automatically or relatively mindlessly; the perceiver has to be committed to this form of
processing. Personalization also makes use of categories but in a different sort of way. In the previous case I may
see Mai Ling as a good representative of my category of Asian student, but she will be seen primarily as just that
and no more unless she is individuated , in wh ich cas e she is st ill an As ian stu dent b ut on e who is clearly
identified as not fitting that category perfectly, perhaps because she is boisterous. On the other hand, as I
personalize Mai L ing (ass umin g I hav e mad e the com mitme nt to u nderta ke the exte nsive p rocess ing requ ired), I
will also see h er in terms of sev eral categories, b ut those c ategories w ill be subordin ate to her. 17 I see her as
female, Asian (in this case Chinese), a graduate student in geology, a person with a good sense of humor, and
someone who is unattractive and obese. These will be attributes of Mai Ling, rather than her being representative
of each of these catego ries. Not only is the representation “flipped” (in the sen se of categories nested un der
17
Category-based processing is close to what we will call intergroup processing and personalization
to what is often ca lled interpersonal processing (Brewer, 199 8). In the former case on e responds to
another (typically an out-group member_ primarily in terms of category membership. In the latter, these
categoriza tions are less relev ant (See Ch apters 6 an d 10 an d p. ?).
Chapter 4: Schemas / Version of November 10, 2002
28
person rather than the reverse) but the form of the representation is also likely to be semantic and propositional
rather th an pict oliteral. See Figure 4 .2.
Fiske and Neuberg’s Continuum Model: Susan Fiske and her students (Fiske, Lin, & Neuberg, 1999;
Fiske & Neuberg, 1990) have argued that there is a continuum of processes used in forming impressions of
individuals. At the one end we have category-based processing; impressions are formed almost entirely on the
basis of stereotypes about groups to which the person belongs. At the opposite end there is attribute-oriented or
piecemeal processing where the impression is based largely on combining behavioral and other individuating
information tha t people display. Fiske and N euberg (1990) argu e that when w e meet a person for the first time,
we try to fit that person into on e of our ready-ma de categories. If her or his behavio rs are consistent with
inferences we might draw from the category itself, then we are likely to form a category-based impression. If, on
the other h and, the p erson disp lays som e attributes th at violate ou r expectancies about h ow a pers on of this s ort
ough t to behav e, then we w ill move to a more sop histicated fo rm of an alysis. We m ight reinterpret the b ehaviors
so tha t they fit th e stereoty pe. For exa mple, a p erson w ho th inks th at wo men a re careless drivers c an still
interpret the driving of that woman who is carefully edging into traffic on the freeway as "much too cautious
and likely to cause an accident." We might attribute inconsistent behavior to luck or other external factors.
Sometimes w e make use of va rious subtypes. Th e articulate black president of the local PT A may co me to be seen
as an exemplar of a black businesswoman rather than as a black person in general for a perceiver who thinks that
blacks can't th ink clea rly or talk intelligently . Failing c ognitiv e devices such as thes e, the perce iver ma y simply
have to pus h all her stereotypes aside and deal with this person w ho violates all her expectancies as a separate
case an d form a n impressio n witho ut benefit of h er stereotypes. Th e net result is tha t to the extent perc eivers
individ uate th ey are m ore likely to pay att ention to inco nsisten t inform ation.
Motivational Factors : This kind of information processing model can often seem to imply an automatic or
machine like series of steps that the perceiver goes through. However, Fiske and Neuberg argue that the
perceiver has a certain amount of control over attention and interpretation through motivational factors that may
enter in at va rious stag es. For exam ple, you mig ht hav e some reas on to w ant to form an accu rate impressio n of a
target person. Perhaps yo u are going to h ire her for a crucial job. This wou ld force you to pay c loser attention to
attributes which would generally push you closer to the attribute-processing end of the continuum. It is also
likely that as you pay closer attention to the person's attributes you will discover more ways that she is not like
your stereotype of people like her. This too w ould force you to consider her even m ore carefully, and encou rage
you even furth er to continue attribute proces sing. Howev er, here as elsewhere, the perceiver may confron t a
classic ability-motivation distinction. All the best intentions will not produce results without requisite ability.
Even people wh o want to ind ividuate do no t do so wh en they do no t have the cogn itive resources because they
are too bu sy (Biesan z, Neub erg, Smith, A sher, & Jud ice, 2001; P endry & M acrae, 199 4).
There is eviden ce for the existenc e of these two kinds of pro cesses. Fiske, N euberg, Bea ttie, and Milberg
(1987) had subjects form impressions of a target based on labels and attributes. When the attributes were not
salient subjects used labels to form impressions, and when there was no label impressions were formed on the
basis of attrib utes. Mo re interestingly, wh en labels an d attributes were incon sistent, impress ions were m ore
attribute-based than when they w ere consistent. Pavelcha k (1989) had subjects evaluate people w ho were defined
by several traits. When subjects were asked to guess the person's academic major and to assess how typical the
person was of that major (conditions that should encourage category-based processing), final evaluations of the
target were better predicted by the subjects' evaluation of the major than by their evaluations of the individual
traits. In another condition, subjects did not overtly categorize the targets, and here the final evaluation was
predicted by eva luation s of the d efining a ttributes , so cate gorica l inform ation w as less im portan t.
Fiske and von Hend y (1992) argue that norms can a ffect whether people are motivated to individuate. In
one study subjects were told that they were particularly good at individuation or category-based judgment. As
predicted subject s who were low in self-mo nitorin g (people w ho gen erally pay attentio n to int ernal cu es), paid
more attention to inc onsistent informa tion when th ey thought th ey were good ind ividuators than when they
thou ght th ey were c ategor y-judg ment ty pes. By c ontra st thos e who are high in self-mo nitorin g were m ore likely
to indiv iduat e when norm s sugg ested th at it wa s appro priate to d o so.
Dependency: According to F iske outcome d ependency is one im portant motiva tor of attribute-based
process ing. W hen w e are depe nden t on ot hers, w e shou ld be mo tivated to predic t their beh avior a ccura tely
because we m ay need to con trol it. Neuberg and Fiske (198 7) had sub jects interact with a suppos ed former
mental pa tient diagn osed as schizop hrenic. W hen their ou tcomes w ere dependen t on him their impression s were
less affected by stereotypes of schizophrenia than when they were not so dependent. Erber and Fiske (1984) had
Chapter 4: Schemas / Version of November 10, 2002
29
subjects work together on a task and some of the subjects thought that the teams that worked together best might
win a prize. In th is cond ition w here the s ubjects ' outco mes w ere partially depend ent on their kno wledg e of their
partners, they were much more likely to attend to the partner's attributes that were inconsistent with previous
expectations. Under so me conditions people also are more likely to individua te those with wh om they expect to
compete (Ruscher & Fiske, 1990; Sanitioso, Freud, & Lee, 1996). It makes sense that one would want as much
diagnostic inform ation as possible ab out a com petitor, but when on e is competing as part of a team, opponen ts
are individuated less than fellow team members because knowledge about team-mates is important for success
(Ruscher, Fiske, Miki, & van Manen, 1991). On the other hand, non-competitive groups that are consensus
oriented tend to focus on stereoty pe consisten t informatio n (Rus cher, Ham mer, & Ha mmer, 19 96).
Power: Power is in many respects the opposite of dependency, and Fiske (1993) has extended these
arguments to say that generally people w ho are powerfu l will be more likely to stereotype those with less pow er
than the reverse. So bo sses will tend to see their secretaries in stereotypic ways, w hereas secretaries are likely to
have a much more differentiated view of their bosses. She argues that those with less power are more dependent
and m ust pay c lose attention to get an a ccurate fix on the pow erful person’s ways o f thinking w hereas the m ore
powerful person o ften needs only to kn ow that the less po werful will behave acco rding to their assigned roles
and du ties. Furthermore where as a given the powerless m ay have on ly one or two peop le who can a ffect them
thus allowing them to devo te more attention to each , the more powerful person might have sev eral subordinates
who need monitoring leaving less cognitive capacity for individuation. On the other hand, while less powerful
people may be highly motivated to individuate the more powerful, they may also be more anxious which may
reduce their cognitive capacity and ability to do the extra work involved (Fiske & Morling, 1996; Fiske, Morling,
& Stevens, 1996). In general, however, according to the Fiske analysis powerful people should pay less attention
to individuating information and when they do attend to such information to pay closer attention to stereotype
consistent as opposed to inconsistent information (Goodwin, Gubin, Fiske, & Yzerbyt, 2000). They may also pay
particular attention to negative information about subordinates as a way of justifying their superior
status(R odríguez -Bailón, M oya, & Y zerbyt, 200 0).
This analysis of power makes a certain amount of sense, and as we will see power often does play an
important role in stereotyping an d prejudice. Howev er, such effects of power are far from inev itable. As a faculty
member, I often do treat my students as an undifferentiated blob of sorts, particularly in large classes or when I
prepare lectures and exams or am cross about some student behavior I don’t like. But at other times I am struck
by what seem s to me to be clear evidence th at I have a lot better sense of h ow my stu dents differ than they seem
to have of how faculty members differ. Surely there are circumstances in which norms associated with particular
power positions or other circumstances lead the more powerful to be especially attentive to the individual
characteristics of their subord inates. In any even t, when pow er is defined in different ways it can lead to grea ter
and n ot less in dividu ation (O verbeck & Pa rk, 200 1).
Accuracy: Goals to be accu rate, often increased by ou r desire not to be embarrass ed by our inacc urate
judgments, can also lead to less reliance on category-based stereotypes (Nelson, Acker, & Manis, 1996; Neuberg,
1989; Neuberg & Fiske, 1987; Ruscher & Duval, 1998). Arie Kruglanski's (1989a; Kruglanski & Webster, 1996)
theory of lay epis temolo gy stress es the ro le of accu racy d eman ds in pe rceiving others . He arg ues th at peop le
generate hypotheses from their theories and then test these hypotheses more or less systematically. The
hypothesis will be accepted as correct if the person cannot easily generate alternative explanations. The
motivation to generate these alternative hypotheses is a function of: (1) the need for structure, (2) fear of
invalid ity, and (3) the n eed for sp ecific con clusio ns. Ne ed for stru cture ca n be inc reased by va rious fa ctors, b ut is
thou ght to b e especia lly sensitiv e to time p ressure and th e need to reach a decision quickly . Therefo re, people
who are under such time pressures may fail to generate alternative explanations and may accept their provisional
hypoth eses unreflectiv ely. Fear of inva lidity is based o n the soc ial costs of m aking m istakes. Peo ple who a re
afraid of such errors will try to consider as much evidence as reflectively as possible. Finally, some people may
have a vested interest in one or another hypotheses. For example, prejudiced people may want to believe that
Hispanics are lazy and hence be relatively unwilling to entertain alternative explanations of behavior that seems
to be lazy.
In one experimental test Kruglanski and Freund (1983) manipulated time pressure and the possibility of
public exposure for errors and asked Israeli teachers to read an essay ostensibly written by a high status
(Ashkenazy Jews) or low status (Sepharadic Jews) ethnic group. As expected the teachers generally rated the
essay more favorably when it was supposedly authored by the higher status student. However, this tendency
was especially pronounced when they were under time pressure and when their fear of exposure for having
Chapter 4: Schemas / Version of November 10, 2002
30
made a m istake was low . These data then s uggest that people m ay make dec isions based on their stereotypes
(initial hypoth eses) as opp osed to b ehaviora l informatio n when they mu st make q uick decisio ns and when th ere
are relatively few social costs to doing do.
Self-Categorization : Although not a part of the formal Fiske model, one’s own standing relevant to the
target often plays a major role in processing concerns. As Self-Categorization Theory (see Chapter 3) argues that
when w e see others in terms of out-grou ps, groups to wh ich we do no t belong, there is a tendency bo th to see
members of tha t group as different from on e’s own bu t also to tend to see them as a more ho mogenized b undle.
By con trast wh en one is perc eiving fellow in -group m embers w here person al and n ot group identity is mo re
salient or when one interacts with others just as people without regard to their group memberships then our
stereotypes should n ot play a major role in ou r perceptions. Thus, wh en self-other relationships are categorized
in terms of groups, perceptions should be less individualized than when the relationships are more interpersonal
and n ot grou p-based (Reyn olds & Oak es, 200 0).
Comparisons of Models: Both Brewer an d Fiske-Neuberg as sume that th e perceiver begins by trying to
fit a new person into an existing category and if successful the target will be seen largely in stereotypic terms.
Both a ssum e that even if the person d oes not rea dily fit an existing c ategory, altern ative wa ys to create a fit are
often used – sub -typing, reinterpretation of behavior, etc. Both ag ree that when the perso n is so motiva ted either
by a lack of fit with readily available categories or because of some extraneous motivation, fairly effortful
process ing can be used to see th e target a s som ething o ther tha n a m ember o f a stereo typed c ategor y.
Beyond thes e important agreemen ts there are also some sa lient disagreements (see Fiske, 1988; Fiske &
Neuberg, 198 9). First, whereas Fiske-Neuberg a rgues explicitly that people first try to use category based
stereotype information and individualize only if that fails, Brewer assumes such a process only if an initial
decision is made to use category-based processing rather than personalization. Secon d, the continuu m mod el
argues that there is no warrant for assuming different forms of representation of people and categories at
different stages of processing wh ereas Brewer thinks tha t information is represented differently. The FiskeNeuberg model also tends to assume that both motivational factors and degree of fit are important at each stage
of processing wh ereas Brewer’s dua l processing mod el tends to use different processing rules at ea ch stage.
While both models recognize an important role for motivationally dictated control over processing, the
contin uum mod el emph asizes this far m ore.
Parallel Processing Accounts
Standard information processing accounts of human cognition emphasize the ways people are like
computers. In pa rticular, the computer I am using to write this cha pter can basically only d o one thing at a time;
it must process inform ation in a certain order albeit so rap idly that I think it is doing m any things a t once.
Similarly both the Brewer and Fiske-Neuberg models assume that we form impressions of others via a series of
quite dis tinct pro cesses. O bviou sly suc h mo dels allo w som e back -trackin g and feedbac k, so th at if a go od fit is
not obtained a t some level the person ma y back up a s tep or two. Still, the person does one th ing at a time.
More recent emphasis on parallel processing both in the computer world and in cognitive psychology
allow for the possibility that what appear to be separate steps can actually be performed at the same time. Read
and M iller (1993) an d Kun da and Thaga rd (1996 ) have pro posed pa rallel processing models fo r how w e form
impressions; we w ill rely on the Kunda-T hagard version . While the details of the mod el are somewha t complex
the ou tline is clea r. In this mod el stereoty pes, trait, a nd beh aviors (as well a s other inform ation) a re all
represented as nodes with varying degrees of connection. So a node representing college professors may have
direct links to traits such as h elpful, lazy, absent-minded, intelligent, arrogant, an d liberal as well as links to
classes of behavior such as lectures and particular behaviors such as talked with a graduate student on Friday.
Other links are indirect. Keeping regular office hours may be linked positively to helpful but negatively to lazy.
Keeping office hours might also be linked to additional traits that are not part of the professor stereotype such as
responsible and friendly. Respo nsible, for its part, may be linked to add itional behaviors su ch as “helps so n with
homework” which in turn may be linked to information such as “helps daughter with science fair project” which
in turn may be linked to other traits, behaviors, and categories (e.g., father). Depending on the strength of the
links and wh ich nodes are ac tivated by variou s contingencies, on e can travel fairly far along these path s or not.
Note th at with in the K und a-Tha gard (1 996) m odel, tra its, categ ories, an d beha viors h ave ess entially
equal status. Th erefore one can mo ve from traits to behav iors or from behav iors to categories with equa l facility
(assu ming t he links are equa lly stron g). Thu s, whe n suc h a m odel is a pplied to t he Lo cksley p aradig m, there is
no a priori reason to b elieve that predictions of behavior sh ould be stronger from labels or from concrete
behav iors. No te also th at the v ariou s links a re travers ed in pa rallel fash ion. Th e perceiver d oes no t have to wa it
Chapter 4: Schemas / Version of November 10, 2002
31
for a given behavior to be linked to one trait and then another; so an assertive behavior may simultaneously lead
to nodes representing
assertiveness, rudeness,
gender, the given
situation. The mo del
also allows for links
both to facilitate and
inhibit activation. So
they may be
strengthened or
weakened as a result of
other information being
considered since each
node is capable of
receiving input from
several other nodes at
the same time. As a
result stereotyp es are
not see n as pr imarily
associations b etween
categories and featu res
but ra ther as d ynam ic
patterns of features and
behaviors that may
change from time to
time, situation to
situation, depending on
which links are
activa ted mo st stron gly.
In this mod el
the
meaning
o f traits
Figure 4.3 : Account of game playing and occupations as cues for intelligence.
and beh aviors is v ery
much th e product of wh at associations are activated, and expecta tions abou t behaviors are a prod uct of other
information such as traits. So, for example, if I ask you to imagine a young janitor and an older professor playing
computer games, you would probably imagine quite different games being played. You might assume that the
janitor would be playing solitaire or perhaps Duke Nukem whereas the professor would be playing some
intricate role-playing game or perhaps one involving sophisticated logic. Now let us suppose further that your
stereotypes of professors includes a link to the node “intelligent” whereas your stereotype of janitor does not. If
you observe both the professor and the janitor playing solitaire, your ratings of their respective intelligence might
not be m uch a ffected. In this ca se playin g solitaire is not es pecially d iagno stic of int elligence s ince nea rly
everyone, smart an d stupid, can a nd does play th e occasional ga me (compu ter based or otherwise). In oth er
words you might rate the professor as more intelligent than the janitor, because links between the two person
types and intelligent are stronger tha n the links between the b ehavior and the trait. This case where stereotypes
are more important than individuating information is diagramed in the top half of Figure 4.3.
Now co nsider the alternative possibility. You o bserve both playing a n intricate game that inv olves
simulating the growth and development of western civilization. In this case the behavior is highly diagnostic of
intelligence, and the links between the beh avior and intelligence are likely to be stronger tha n the links between
professors or janitors and intelligence. This then would represent a case where individuating information seems
to swamp stereotypes. This is diagramed in the bottom half of Figure 4.3.
Finally w e note th at stereo types so metim es help u s interpret ambig uou s beha vior. Im agine o f you w ill
that you observe a visitor to campus asking directions. She asks directions of a custodial worker or a person who
appears to be a professor, and in both cases the target shrugs his shoulders and walks away without giving
directions. In the case of the cu stodial person you might have links to traits such as u nintelligent, unfamiliar with
campus, or does not speak English well as well as potential links to traits such as lazy and unhelpful. For the
Chapter 4: Schemas / Version of November 10, 2002
32
profess or thes e first links w ill not be pr esent (a nd ind eed will pro bably b e reversed ) so tha t they ca nnot explain
the behavior. On the other hand, if your stereotype of professor includes links to traits such as “head in the
clouds” these will provide reasonable ways to interpret the behavior. So the same behavior may be interpreted as
due to stupidity or lack of knowledge on the part of the grounds keeper and as rudeness on the part of the
professor.
One major advantage of this type of model is that it erases what turns out to be a problematic distinction
between category and individuating/behavioral information. It also obviates the need for debate about different
sorts o f process es, their seq uenc ing, an d differen t mod es of repres enting in forma tion. In that sen se the m odel is
simpler than the sequential models of Brewer and Fiske and Neuberg. What has been left out, however, may be
fairly important. There is little room for the sorts of motivational factors we have been discussing. Further since
much of this processing is assumed to take place outside of awareness, it is unclear to what extent people can
exert control over how they consider information about others. Finally, this model does not readily allow for
different forms of representation of groups and individuals or for the effects of emotion on processing
information about others. To these two areas we now turn.
On-Line vs. Memory-Bas ed Processing
The d istinctio ns betw een sequ ential-ser ial proces sing an d para llel process ing disc ussed in the las t section really
refer mostly to h ow ou r mental ap paratus deals with informatio n once it h as been en coun tered. How ever, there
is a closely related issue of how we deal with inform ation that we get seriatim. One of the classic issues in person
perception research has b een whether we respo nd differently to information w e acquire early as opposed to late
about a person (Schneider, Hastorf, & Ellsworth, 1979). Obviously, since we can never find out everything we
know about others simultaneously, we take in relevant information a little at a time. The impressions you have
formed of even close friends o r family members ha ve been built up over time.
The Hastie-Park Analysis: An influential paper by Ha stie and Park (19 86) dealt with just this issu e. They
attempted to resolve a paradox of sorts. Often it seems the favorability of the impressions we have of others bear
little relation ship to wha t we ca n easily remem ber abo ut them . It wou ld seem reason able tha t if I recall m ostly
positive things abou t a perso n I ha ve just m et that I w ould h ave a p ositive im pression of that person . Yet, often it
seems memories and evaluations are not closely related. Hastie and Park argued there are two ways we might
approach information as it comes in. In what they called on-line processing, one considers each piece of
information and uses it to update a developing impression. Thus the impression is constantly being revised. In
such on-line processing we would normally expect that information encountered first would be most important
in contributing to the o verall impression, and su ch primacy effects are frequently foun d. Howev er, when asked
to recall information about the person information that occurs later (recency) might dominate. In the case of online processing it is hard to predict wh at relationship (if any) will be found b etween memo ry and judgm ent.
The second mode of process ing is memory-ba sed. In this case, the perceiver does no t actively integrate
each new piece of information into an over-all impression. Indeed, the perceiver may not even have a goal of
forming an impression. In that case, when an impression is called for there is none to be had, and it must be
formed on the spot fro m wh at can be remem bered a bout the targ et. Thu s there w ill be a clos e relation ship
between what is remembered and the final impression.
Impressions of Groups and Individuals: Hamilton and Sherman (1996; Hamilton, Sherman, & M addox,
1999) have used this distinction to argue that groups and individuals will be perceived in different sorts of ways.
They argue tha t when we perceive peo ple we generally assum e that each individu al has a core con sistency to her
or his personality. Thus, to the extent we care about the target, we strive to develop an organized, consistent
impression. This in turn biases us toward processing information about a person in an on-line fashion. When we
process information about individuals on-line, we would expect to see primacy effects in impressions,
spontaneous inferences about personality based on behavior, and explicit attempts to deal with inconsistent
informatio n (often resu lting in better mem ory for the inc onsisten t informatio n).
I meet Jane for the first time and immediately notice that she is stylishly dressed. Because I assume that
she has con sciously chos en to dress in a particular w ay, I immediately decide tha t she is a person wh o cares
about appea rances. Perhaps sh e is superficial, but I’ll withhold judgmen t on that for now . Next I hear her mak e a
comment about a lengthy lunch she has just had with “the ladies” at a local expensive restaurant. This piece of
information confirms my suspicion that she cares about appearances and tilts me toward the further inference
that she is s uperficial, con cerned on ly with imp ressing peop le. But then I h ear her ma ke an intelligent rem ark
about a loca l art show, and I begin to won der. Perhaps she is no t so superficial after all. I hear her ask in wha t I
take to be a medically sophisticated and sincere question about the illness of a mutual friend, and I’m now
Chapter 4: Schemas / Version of November 10, 2002
33
beginning to get the sense that there may be more to her than I thought. And so it goes. Each new piece of
information is given a larger meaning in terms of dispositions and motives, and I try to make each new piece of
information fit into my developing impression. I expect Jane to be of a piece, and I’m working to make my
impression coheren t.
How ever, no rmally, w e do no t expect th at grou ps will h ave th e same degree o f unity a nd co herenc e. I
observe a grou p of well-dressed wom en talking abou t their lunches and artistic tastes, and I may o r may not n ote
basic agreements. But even if they are all praising the same artist, I do not necessarily assume that they are doing
so as mem bers of a group. I wo uld expect Jane to be con sistent from mom ent to mom ent, conversation to
conversation in her liking for a particular artist, and if she is not I am likely to try to figure out why she is being
so inconsistent. However, when I observe a group of women who express differing opinions, I am not the least
bit surprised because I have no reason to believe that they should agree. Because we do not have the sense that
groups have a coherent “personality”, we are not as likely to process information about them on-line. Thus, our
impress ions o f group s are m ore likely to be mem ory-ba sed.
Entitativy: How ever, Ham ilton and Sherma n (1996 ) recognize th at we expect s ome gro ups to h ave mo re
unity than others.18 While a gro up of wo men talkin g abou t art may h ave relatively little in co mmo n, memb ers of a
family, a religious cult, or a fraternity, may often seem to be more tightly knit together and to share more common
features. Ca mpbell (195 8) has a rgued so me grou ps seem to be entities in som e psycho logical sens e whereas others
are less so, and Hamilton and Sherman suggest that for groups with high entitativy we expect a fair amount of
consistency am ong mem bers and that w e are likely to use on-line processing. Grou ps that are not perceived to b e a
distinc t and c oheren t entity w ill be appro ached in a me mory -based man ner.
Several implications of this mo del have been tested. Keep in m ind that the ma in difference between
perceptions of groups and individuals is the natural and normal tendency to use memory-based processing for
groups and on-line for individuals. Ho wever, as McC onnell, Sherman , and Ham ilton (1994a) ha ve shown when
perceivers approach bo th groups an d individuals w ith a goal to form an impression, information about both
should be processed in an on-line manner. Conversely goals that direct attention away from impression formation
lead to memo ry-based processing for bo th. McCon nell, Sherman an d Ham ilton (1997) directly man ipulated
expectations of entitativity for both groups and individuals. Subjects were given information about an individual
or a group an d were told that the ind ividual was either co nsistent or not an d that the grou p members w ere either
highly similar or not. As expected when the group or individual was high in entitativity, information was
processed in an on-line manner but when low processing tended to be memory-based. Thus while we might
generally process information about groups in a memory-based way and about individuals in an on-line fashion,
these natural tendencies can be overcome either by changing motivation or by altering assumptions about the
relative entitativity a bout gro ups an d individu als. Becau se people genera lly think tha t individu als are mo re “of a
who le” tha n grou ps. The y mak e more ex treme tra it judgm ents ab out in dividu als an d wo rk hard er to reco ncile
inconsistencies in their impressions of them (Susskind, Mauer, Thakkar, Hamilton, & Sherman, 1999)
Brewer and Harasty (1996) have argued that the perceived entitativity of groups may also be stronger for
out-groups th an for in-groups (see Ch apter 6), and that w hen social identity is crucial even in -groups may be seen
as having s trong entity properties (see also Mullen, 199 1). Perceptions of group en titativity may also be increased
by a va riety of goals a nd oth er motives (B rewer & H arasty, 19 96). For exa mple, durin g freshma n orientatio n every
fall semester, I am struck by how easy it is to divide the world of students into freshmen and upperclassmen and
to thin k of the fo rmer as being a distinc t group with th eir own needs a nd bu rgeonin g cultu re.
Implications of Entitativity : Assuming that category-based processing (Fiske & Neuberg, 1990) should be
largely m emory -based , where m ore ind ividua ted proc essing s hou ld be mo re on-line , perceived entitativ ity sho uld
also affect how much attention we give to the two kinds of information. For example, our perceptions of groups
may be largely based on stereotype information whereas perceptions o f individuals from th ose groups m ay have a
18
Indiv iduals can a lso va ry in their p erceived cons istency . Interestin gly we h ave a v ariety of tr ait
attributions for those who are inconsistent. Sometimes we see them as moody. People who express
different opinions to different people may be seen as insincere or hypoc ritical. In extreme cases
unpredictability may lead to perceptions of mental illness (See Chapter 9). It is striking that we do not
have such a rich vocabulary to describe inconsistencies among group members presumably because we
expect groups to be somewhat inconsistent. There is probably also less variability in our perceptions of
individual than of group inconsistency.
Chapter 4: Schemas / Version of November 10, 2002
34
fighting chance to be more individuated. Mullen (1991) has argued that we can represent groups in terms of
prototypes or exemplars, an d it is reasonable to believe that grou ps high in perceived entitativity are largely seen
in terms of pro totypes (Brew er & Hara sty, 1996 ).
Another implication is that groups high in entitativity should also b e high in inductive po tential (See
Chapter 3). That is, just as we assume that one or two instances of behavior on the part of an individual give us the
cognitive right to assu me that similar beha viors will appear in the future, so grou ps that are high in entitativity
will also encourage us to assume that we can generalize the behavior of one person in the groups to others. In a
related argument Hilton and von Hippel (1990) have shown that for groups high in entitativity, individuals who
are moderately discrepant from the group mean are assimilated to the group stereotype whereas for less
homogeneous groups such individuals are contrasted from the stereotype. This is, of course, the very essence of
stereoty ping -- it is im portan t to get a b etter gras p on h ow w e perceive g roups to differ in t his rega rd.
Affect and Mood
Cognitive psychology generally and social cognition in particular have traditionally avoided discussion
of mood and affect in information processing accounts. However, there is now abundant evidence to suggest that
how we feel affects how we think (as well as the reverse). This is likely to be especially important in stereotyping
because stereotypes themselves are affectively laden and have complex connections to prejudice and
discrimination, themselves affectively driven processes. An obvious prediction is that our moods should affect
the particular content of our stereotypes, but the effects of affect, mood, and emotion also affect the ways we
process inform ation (s ee Forga s, 199 5, and Wyer, C lore, & Is bell, 199 9, for recen t reviews ).
The Effects of Mood on Stereotype Content: Following such models perhaps the most obvious and
comm on-sen sical pred iction th at one might make is that n egative m oods shou ld prom ote stereo typing e specially
of groups that are negatively evaluated. Certainly there are abundant data suggesting that people in a good
mood are more generous in their evaluations of people and things than those in a bad mood (Bower, 1991; Forgas
& Mo ylan, 1 991). S uch m ood-c ongru ent effects have b een repo rted in th e stereoty pe area a s well. Fo r examp le,
Esses and Zanna (1995) had subjects list attributes of several ethnic groups after a positive, negative, or no mood
was ind uced; follow ing this they rated the po sitiveness o f the features th ey had listed . For those g roups th at were
the most nega tively evaluated (native Ind ians, Pakistanis, an d Arabs) people in a neg ative mood ten ded to rate
the features they had listed more negatively, and other experiments suggested that the mood affected how the
features were rated more than what features were actually listed. Haddock, Zanna, and Esses (1994) also found
that subjects in a pos itive mood ha d more positive stereotypes an d more positive emo tional association s to
stereotyped groups than those in a negative mood.
There are two major ways we might explain such effects. One of the earliest ideas in the affect-cognition
literature was that affect becomes associatively linked to other content in our memories and hence can serve as a
retrieval cue for affectively congruent material (e.g., Bower, 1981; Isen, 1984). During encoding mood may also
affect what information we pay attention to and what information gets encoded. Mood may also affect how
inform ation g ets interpr eted. So when people in a negativ e moo d stay focus ed on the even ts asso ciated w ith their
moods they remained in that state, whereas those who try to reappraise the negative events in a more positive
way (R usting & DeH art. 200 0).
Schwarz, Clore and their colleagues (Clore & Parrott, 1991; Schwarz, 1990; Schwarz & Clore, 1988;
Schwarz & Bless, 1991) ha ve proposed an alterative view. They argue th at mood s are used as inform ation when
making judgments, and that we often mis-attribute their moods to the stimulus being judged. Thus a person in a
bad m ood m ight rate a grou p more negativ ely than a perso n wh o is in a g ood m ood b ecaus e she im plicitly
assumes that the groups themselves are causing her affective state. On the other hand, if mood is attributed to a
judgment relevant state, there is no need for further (mis)attribution, and mood should not affect stimulus
ratings, a prediction that has been confirmed (Schwarz & Clore, 1988). If this account is correct then moodindu cing ev ents m ay also serve as stand ards o f reference a nd if extrem e enou gh lead to con trast effect s. Abele
and Gendolla (1999) have shown that after positive or negative moods have been induced important aspects of
life (presumably extremely affective) are judged as more positive after negative than positive moods with the
reverse for negative moods. Less important (and extreme) aspects of life showed assimilation to mood.
Mood and Processing Information: Moods and affect also affect how we process information about
others. Earlier in this chapter we no ted that stereotypes often act as heuristics and tha t they will be most often
Chapter 4: Schemas / Version of November 10, 2002
35
used when people are not motivated to process information about another person systematically. Mood may
affect our tendencies to process information systematically or heuristically.19
Moods Affect the Use of Stereotypes: Bodenhau sen and h is colleagues have m anipulated mo od and th en
asked subjects to jud ge either a cheating or an a ssault infraction w here the offender is stereotypically linked to
the infraction (i.e., an Hispanic defendant for an assault case and an athlete for cheating). Whereas subjects in a
sad or neutral mood do not rate the stereotypic defendant as more guilty, those in a happy mood do
(Bodenhausen & Kramer, 1990; Bodenhausen, Kramer, & Süsser, 1994). Subjects who are angry (Bodenhausen &
Kramer, 1990; Bodenhausen, Sheppard, & Kramer, 1994), anxious (Wilder, 1993; Wilder & Shapiro, 1989a;
1989b)or stressed (Keinan, Friedland, & E van-Ha im, 2000) also m ake more stereotypic judgm ents. Other studies
(e.g., Krauth-Gruber & Ric, 2000; Mackie, Hamilton, Schroth, Carlisle, Gersho, Memeses, Nedler, & Reichel, 1989;
Park & Bana ji, 2000;) a lso sho w tha t people in a goo d mo od are especially prone to use stere otypic th inking .
Thus people who are happy or angry (but not sad) seem more likely to use their stereotypes as heuristic devices.
Why Affect Influences Heuristic Thinking: There may be ma ny reasons for th is. One possibility is that when
our moods are good we are generally inclined to slack off a little, to be just a little lazy. It is as if people reason:
Well, my mood is fine; hard work can’t improve it and might lead to frustration -- let’s not rock the boat. Such a
model would explain why happy people use stereotypes more than sad people do, but not why anger also seems
to produce stereotypic thinking.
A second po ssibility is that when we are in a good mo od, we may also be distracted with h appy thoug hts
and or otherwise have less capacity to process complex material (Bodenhausen, 1993; Mackie & Worth, 1991;
Wilder & Simon, 1996). Angry thoughts may also be distracting and reduce the capacity for systematic thinking.
By contrast people who are sad are inclined to be especially systematic in their processing. It is unclear why sad
people a re not a s distra cted by emotio n-related thou ghts a lthou gh it is po ssible th at wh en we a re sad w e actively
avoid thinkin g abo ut neg ative ide as. Do ubt h as been cast o n this a ccou nt beca use of ev idence t hat h appy pe ople
do no t show diminish ed cogn itive capacity a s predicted (Bles s, Clore, Sch warz, G olisano, R abe, & W ölk, 1996 ).
A third account (Schwarz & Clore, 1983; Schwarz, 1990) is based on the idea that moods provide
information. Good moods suggest that the world is fine and needs no special cognitive work whereas sad moods
may act as a cue that all is not right w ith our cognitive w orlds and w e need to restore understa nding and closure.
Howev er, we should no t assume tha t all negative mood s work the sam e way. Anger an d anxiety tend to be a
more in tense em otion t han s adne ss, and may fin d proc essing in forma tion ca refully esp ecially diffic ult. Th is
would explain why angry and anxious people are more likely to use their stereotypes as aids in processing
information. Furthermore more recent versions of this model, now labeled the Mood and General Knowledge
Mod el (Bless, Sc hwa rz, & K emme lmeier, 19 96) ha ve sug gested th at goo d mo ods en coura ge heu ristic thin king in
that such moods signal that existing interpretations and cognitive structures are right with the world, whereas
sad moods lead to an active search for information about why the world seems so negative. Hence those in a
positive moo d sho uld rely o n stereo types m ore an d tend to igno re beha vioral o r other in forma tion th at is sligh tly
inconsistent or ambiguous. However, even good mood people are not likely to ignore blatantly inconsistent
information, and will use their stereotypes less and the inconsistent behavioral information more. An experiment
by Kra uth-G ruber a nd R ic (200 0) con firm thes e predictio ns.
Affect and Processing Information About Groups: Affect has effects other than simple use or non-use of
stereotypes. For example, people in good moods tend to use broader and more inclusive categories (Isen &
Draub man, 1 984) at leas t when th ey focus o n similarities am ong exem plars (Mu rray, Sujan , Hirt, & Sujan , 1990).
Sinclair and Mark (1992) have suggested that people in good moods adopt more flexible strategies in dealing
with information , using broad ca tegories when this w orks and is easy but using m ore narrow ca tegories when
necess ary. Fied ler (1988 ) sugg ests tha t positiv e moo ds enc oura ge a mo re intuitiv e and flexible cog nitive sty le
(called loosening) whereas negative moods promote a more systematic, analytic style. People in good moods also
tend to pay less attention to information ab out how variable groups are tha n those in neu tral moods (Stroess ner
& Ma ckie, 199 2) and also ten d to dis crimina te more b etween in-grou ps and out-gro ups (A bele, Gen dolla, &
Petzold, 1998). People in a good mood tend to favor the in-group over an out-group more than people are sad or
in a neutral mood provided the relevance of the group is low but sad subjects discriminate more when group
19
There are major similarities and pa rallels between Fiske and N euberg’s (1990) distinc tion between
category-based and attribute-based processing on the one hand and heuristic and systematic processing
on th e other.
Chapter 4: Schemas / Version of November 10, 2002
36
relevance is high (Forgas & Fiedler, 1996). There are several other demonstrations that people in a good mood
process information about others in a less complex way (Seta, Hayes, and Seta,1994; Bless, Clore, Schwarz,
Golisa no, Ra be, & W ölk,19 96; M artin, W ard, Ac hee, an d Wy er,1993 ; Edw ards & Weary , 1993 ).
Finally, Lambert, Khan, Lickel, and Fricke (1997) have shown that people who are sad tend to correct for
their negative but not th eir positive stereotypes. People who w ere in a happy mo od and w ho disliked sororities
had more favorable impressions of non-sorority members on negative traits as did those in a neutral mood.
However, those in a sad mood who disliked sororities actually judged the sorority members more positively on
negative tra its. Lamb ert and h is colleagues argue tha t becaus e people in a sa d moo d are mo tivated to b e more
thoug htful in their pro cessing, they think their stereo types are ina ppropriate ba ses for judgm ents and hence try
to correct (in this case they over-correct). One could argue that such results only say something about mood
congruency effects since sad moods affected only ratings of negative traits, but the argument is more general. In
anoth er experiment, su bjects rated w omen w ho were a ttractive (and for who m they h ave positiv e stereotypes).
When judging them for a job in which attractiveness was said to be relevant sad subjects were more willing than
the neutral mood subjects to hire the attractive candidate, but when attractiveness was said to be irrelevant to the
job, sad subjects were less likely than neutral mood subjects to hire her. Again, this suggests that people in a sad
mood are more likely to process information thoughtfully than those in neutral or happy moods.
Conclusions
There is no simple answer to the questions of whether stereotypes or behavioral and other individuating
information is the m ore important in ou r perceptions of others. Fortuna tely, our mental appara tus is constructed
so there can be no simple answer. Ima gine how aw ful it would be if we were alw ays prisoners of ou r stereotypes
or inevitably had to form new generalizations and expectancies with each new experience. Some of our
stereotypes are strongly entrenched and will resist interference from individuating data. But most are not that
impervious to da ta, and even w ith those that are we h ave the capacity to m entally override stereotypes and to
cons ider peop le as ind ividua ls.
Chapter 4: Schemas / Version of November 10, 2002
37