9781405124003_4_003.

qxd 10/31/07 2:55 PM Page 42

Social Perception and

3 Attribution
Brian Parkinson

K E Y CO N C E P TS
actor–observer difference
analysis of non-common
effects
attributional bias
augmenting principle
causal power
causal schema
central trait
cognitive algebra
configural model
consensus information
consistency information
correspondence bias
correspondent inference
theory
covariation theory
depressive realism
discounting principle
distinctiveness information
false consensus bias
implicit personality theory
learned helplessness theory
naïve scientist model
peripheral trait
primacy effect
salience
self-fulfilling prophecy
self-serving biases
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 43

CHAPTER OUTLINE

How can we tell what other people are like? How do we explain their actions and experiences (and
our own)? This chapter introduces research intended to answer these questions. Studies of social
perception show that impressions of others depend on what information is presented, how it is pre-
sented, and on prior assumptions about how it fits together. Research into attribution demonstrates
that perceivers consistently favour certain kinds of explanation over others. Our impressions and ex-
planations are also shaped by our specific reasons for constructing them. In particular, we present
social events in different ways to different people under different circumstances. Both social percep-
tion and attribution therefore involve communication in combination with private interpretation.

Introduction
Can you remember when you first met your closest friend? How quickly did you get a sense of
what he or she was like, and of how well you would get on together? Did your impression turn out
to be correct, and if not, where and why did you go wrong?
Now imagine that instead of meeting another person face to face, you are told about them by
someone else. When we describe other people, we often refer to their traits (relatively consistent
personality characteristics or abilities) or dispositions. Peculiar as it might seem, let’s suppose that
the only information you are given is the following list of traits:

intelligent – skilful – industrious – warm – determined – practical – cautious

How easily did you form an impression this time? Did you reach your conclusions in the same way
as when you first met your friend? Are you as certain that your judgement is correct?
It is unusual to meet someone without knowing anything about them. Even if you haven’t been
told what to expect, the specific location for your meeting (a bar, concert or supermarket) can be
revealing. You can already tell that they must be the sort of person who goes to a place like this,
and this category information may provide sufficient evidence for your purposes (see Chapter 4, this
volume, and Fiske & Neuberg, 1990). However, we sometimes start with very few clues and need
to construct impressions from scratch. And we often make up our minds about whom we like and
dislike before any conversation begins.
But people we dislike at first can later turn out to be excellent company, and people we think we
will like may ultimately prove less congenial. In any extended relationship, we get to see how the
other person acts in different situations and use these observations to draw conclusions about their
feelings and personality. It is rare indeed that all this subsequent information perfectly matches first
impressions.
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 44

44 CHAPTER 3 SOCIAL PERCEPTION AND ATTRIBUTION

This chapter is about how we make sense of other people. forming impressions of others, and on how the nature of the
Because we draw inferences about someone else’s personality so presented information may also make a difference. The rest of the
readily and usually have little trouble understanding the meaning chapter concerns attribution theory – the study of people’s causal
of their actions, it may seem that our social perceptions are explanations. We present two general models of how information
straightforward and direct. However, the fact that we often have to is processed in order to infer the causes of behaviour. We then
correct initial impressions suggests that things may be more com- consider how attributions can influence our motivations and
plicated. Most social psychologists believe that we piece together emotions, and examine evidence for various biases in attribution.
and weigh up available information before arriving at any conclu- Next we consider the role of language and conversation in deter-
sion, even when we are not explicitly aware of going through the mining attributions and apparent attributional biases. Finally, we
various stages of such a process. And each additional stage brings raise the question of how basic data-driven perceptual processes
another opportunity for bias to creep in. might combine with conversational processes in social perception
The next section of this chapter will review research into and attribution.
social perception, focusing on how information is combined when

SOCIAL PERCEPTION central trait that reconfigures

the meaning of the target’s
central trait a dispositional characteristic
viewed by social perceivers as integral to
whole personality, whereas the organization of personality
politeness is a more peri-
peripheral trait within impression
How do we form impressions of people? pheral trait that has only formation, a trait whose perceived presence
specific and delimited effects. does not significantly change the overall
The contemporary approach to social perception derives from However, Asch found that interpretation of a person’s personality
pioneering research conducted by Asch (1946). What struck Asch trait centrality also depends
was how rapidly we seem to arrive at impressions, despite the on what other words are pre-
diversity of information that has to be combined. How, then, do sented, and that no word is central across all possible contexts (see
we construct a unified picture of someone’s personality from Zanna & Hamilton, 1972).
different pieces of information? Other experiments showed that the order in which adjectives
To investigate this process, Asch read out personality adjectives are presented also made a difference. In particular, earlier informa-
to students and asked them to form an impression of the person tion seemed to exert a disproportionate impact on impressions.
(target) described by these words (just as you were asked to do at For example, a target described as intelligent, industrious, impul-
the beginning of this chapter). Participants wrote a brief descrip- sive, critical, stubborn and envious was seen as competent and
tion of the target and then ticked any relevant traits on a personal- ambitious, but when exactly the same words were presented in
ity checklist (e.g., they had to say whether the target was generous reverse order (so that ‘envious’ came first), the target was thought
or ungenerous, humorous or humourless, and so on). to be overly emotional and socially maladjusted. This greater
One of Asch’s first studies compared two lists of adjectives that influence of initial informa-
were identical except for a single word. The first list contained the tion is generally known as primacy effect the tendency for
same words that you read earlier (intelligent, skilful, industrious, a primacy effect. Evidently, information presented earlier to be more
warm, determined, practical, cautious). Think about these words people do not wait until all influential in social perception and
interpretation
again. Which do you think had the greatest influence on your evidence is in before starting
impression? to integrate it.
The second list simply replaced the word ‘warm’ with the word Kelley (1950) found similar effects on judgements of some-
‘cold’. Asch found that this single change made a big differ- one with whom participants actually had direct contact. A guest
ence. Participants hearing the ‘warm’ list were far more likely to lecturer was introduced to students either as ‘cold’ or ‘warm’ and
describe the target as generous, wise, good-natured, etc. (see students rated him only after he had taught them. Not only was
Figure 3.1). A typical description was: ‘A person who believes the lecturer rated less positively when he had been described as
certain things to be right, wants others to see his point, would ‘cold’ but also students interacted with him less and asked fewer
be sincere in an argument and would like to see his point won’ questions. Since the first thing that students learned about this
(Asch, 1946, p. 263). By contrast, a typical description of the ‘cold lecturer was that he was either warm or cold, these effects could
personality’ was: ‘A rather snobbish person who feels that his depend on either primacy or trait centrality. Further, because most
success and intelligence set him apart from the run-of-the-mill introductions focus on a speaker’s positive rather than negative
individual. Calculating and unsympathetic’ (p. 263). qualities, describing a lecturer as ‘cold’ may have had more impact
In the next experiment, Asch replaced ‘polite’ with ‘blunt’ on ratings and behaviour than under other circumstances.
instead of ‘warm’ with ‘cold’ and found that this change made The results presented so far suggest that people do not simply
much less difference. This suggests that warmth is seen as a add together the bits of information they receive about a target,
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 45

SOCIAL PERCEPTION 45

100 Warm Cold

90

80
Percentage of participants ticking adjective

70

60

50

40

30

20

10

0
Emotional Optimistic Informal Cheerful Short Modest Imaginative Thin Pale

Figure 3.1 Impressions of ‘warm’ and ‘cold’ targets (Asch, 1946).

but rather actively construct meaning based on their ideas about therefore attach more weight to its connotations. Asch’s con-
how different personality characteristics tend to go together. As figural model, by contrast, implies that central adjectives change
later theorists concluded, people have their own implicit person- the meaning of other words rather than simply attracting greater
ality theories that help to make sense of other people (e.g., Bruner emphasis.
& Tagiuri, 1954). More gen- But do social perceivers always make sense of personality
erally, people seem to integ- information in either of these ways? In Asch’s (1946) experiments,
implicit personality theory an integrated
set of ideas held by social perceivers about rate social information by participants heard a list of separate personality adjectives and
how different traits tend to be organized trying to infer its holistic pat- were explicitly told to construct an impression based on these
within a person tern (configural model). words (as in the task at the start of this chapter). As Asch
configural model a holistic approach to The alternative cognitive acknowledged, this is unlike what normally happens when we
impression formation, implying that social algebra model suggests that meet someone face to face (e.g., getting to know your best friend
perceivers actively construct deeper separate pieces of information for the first time). How then might this particular way of pre-
meanings out of the bits of information that are simply added together senting information have affected the process of impression
they receive about other people
or averaged (e.g., Anderson, formation?
cognitive algebra a proposed process for 1981). For example, if a per- Some people get acquainted by email before ever physically
averaging or summing trait information son is described as ‘warm’ but meeting (see Chapter 10, this volume). It may take months or even
when forming impressions of other people
‘boring’, the overall impres- years before they so much as exchange photographs (not neces-
sion would be less positive sarily genuine ones: see Ben-Ze’ev, 2004), assuming that they ever
than if she were described as ‘warm’ but ‘interesting’, but more do (see Joinson, 2003). What happens when these people finally
positive than if she were described as ‘cold’ and ‘boring’. Accord- confront one another in the flesh? Are they surprised by what
ing to this view, the disproportionate effect of ‘central’ adjectives they see?
depends on them conveying comparatively more evaluative in- Examples such as this suggest that transmitting information
formation than the other words that are presented. Further, the in words rather than raw sensory data (sights, sounds and smells)
impact of a word may depend on its relevance to the judgement can make a difference to the content of our impressions. Indeed,
being made. For example, we care more about whether someone sensory information can carry direct implications about person-
is ‘warm’ when selecting a potential friend than a plumber, and ality. For example, people with large, round eyes, short noses, high
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 46

46 CHAPTER 3 SOCIAL PERCEPTION AND ATTRIBUTION

them and keep it to yourself. Instead, you adjust how you behave
to what you think they are like, and they correspondingly adjust
their conduct to their impression of you (which is partly based
on how you are responding to them, and so on). For example, if
you think someone is friendly, you may be more friendly back,
leading them in turn to reciprocate your friendly response (and
so on). Thus, our impressions of others can lead to self-fulfilling
prophecies (e.g., Snyder, 1984). However, people are also able to
adjust their impressions when
expectations are disconfirmed.
self-fulfilling prophecy when an originally
Indeed, if you know that false social belief leads to its own fulfilment.
someone has the wrong idea Social belief refers to people’s expectations
about you, you may deliber- regarding another group of people. When
ately act in ways that show a self-fulfilling prophecy occurs, the
perceiver’s initially false beliefs cause
them that they are mistaken targets to act in ways that objectively
(a self-verification effect, Swann, confirm those beliefs
1984).

SUMMARY
The study of social perception focuses on how we as social
perceivers form impressions of other people, and how we
combine information about them into a coherent over-
all picture. Pioneering studies showed how important
the nature and order of presented information are, and how
perceivers actively construct meaning, rather than simply
sum information. But how specific pieces of information
are weighted, integrated and used depends on a variety of
factors including the situation we find ourselves in, and how
much we care about making the right judgement.

Plate 3.1 People with large, round eyes, short noses, high
foreheads and/or small chins are typically perceived as less
dominant, more naïve and warmer than people with mature- ATTRIBUTION THEORY
seeming features.

What are the main theories of causal attribution, and how do they
foreheads and/or small chins (baby-faced individuals) are typically envisage that lay perceivers process causal information?
perceived as less dominant, more naïve and warmer than people
with mature-seeming features (e.g., Berry & McArthur, 1986), and In one of Pixar™ animation studio’s earliest short films, the move-
people with louder or higher-pitched voices are often perceived ments of two anglepoise desk lamps – one large, one small – are
as more extraverted (e.g., Scherer & Scherer, 1981). The way that accompanied by voice-like sounds. Although items of office furni-
patterns of sensory information change over time can also carry ture do not usually have social relationships, viewers quickly con-
important information. For example, we are quite accurate at judg- clude that the larger lamp is the smaller lamp’s parent, and that
ing which of two people is older by observing the way that they the smaller lamp is a rather boisterous child. The lamps’ contrac-
both walk, even when all other evidence is removed. Adults with tions and extensions soon appear to be actions, and the noises start
a younger-seeming gait are also perceived as more energetic sounding like communications or expressions of emotion. A little
(Montepare & Zebrowitz-McArthur, 1988). drama of conflicting desires and thwarted impulses seems to un-
Asch’s procedure also differs from most everyday interactions fold on the screen, even though we know that every movement
because it provides no opportunity for the other person to respond has been computer-generated. How is this impression of human
to participants’ judgements of them (or for participants to respond personality and intention achieved? Part of the answer is that our
to these responses). Interactivity of this kind may make a big dif- tendency to see motives and dispositions behind human actions
ference to the process of impression formation. For example, when may be so automatic that we sometimes find it hard to override it
you meet someone, you don’t simply draw a conclusion about even in situations where motives and dispositions don’t really
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 47

ATTRIBUTION THEORY 47

apply. Attribution theory (e.g., Försterling, 2001; Heider, 1958; allows us to integrate a variety of otherwise disorganized infor-
Kelley, 1972) provides a set of ideas about how these kinds of mation about others, just as knowing that a larger lamp has a
inferences about the causes of action are made in the more usual maternal attitude to a smaller lamp makes sense of an otherwise
situation of observing or hearing about a human being’s actions baffling piece of animation (see also Heider & Simmel, 1944).
(rather than those of a desk lamp). It addresses our explanations Second, it permits prediction (and, to some extent, control) of
of our own as well as other people’s behaviour. future behaviour. For example, knowing that you are a friendly
Most of the phenomena investigated by attribution researchers person means that I can expect a friendly reaction from you when
involve an observer explaining an actor’s behaviour towards a we meet again.
human or non-human object (or entity), but sometimes the actor
and observer can be the same person (self-attribution). Unlike much
of psychology, attribution research is not directly concerned with Correspondent inference theory
why actors do what they do, but focuses instead on what observers
conclude about why actors do what they do (e.g., whether they How do perceivers decide why one action, rather than others,
attribute behaviour to an actor’s or object’s characteristics or is performed?
‘attributes’). In the parlance of the theory, to make an attribution is
to assign causality to some person, object or situation. According Jones and Davis (1965) tried to make Heider’s ideas about dis-
to attribution theory, we are all amateur psychologists trying to positional attribution more systematic. Like Heider, they argued
explain each other’s behaviour and our own. that observers learn most from actors’ behaviour when it pro-
For example, imagine a friend (actor) has just spent a substan- vided information concerning their personal characteristics.
tial proportion of her student loan on an expensive digital camera For example, you would probably attribute your friend’s camera
with all the latest features (entity). This might lead you (as observer) purchase to her specific intention (buying the camera did not
to think about what provoked such a purchase. Was it an ‘impulse just happen to your friend, she decided to do it), and may in turn
buy’ reflecting a failure of your friend’s self-control? Was she talked attribute this intention to an underlying disposition, such as
into it by a canny sales assistant? Or did her deep-seated interest enthusiasm for photography. Jones and Davis called this process
in photography motivate her spending? Was the camera so spe- of inferring dispositions from
cial that she just had to have it? Or had other friends persuaded behaviour correspondent in- correspondent inference theory proposes
her that she couldn’t do without it? Our answers to these ques- ference because observers infer that observers infer correspondent intentions
tions shape our reactions and our expectations about her future intentions and dispositions and dispositions for observed intentional
behaviour. that correspond to the behavi- behaviour under certain circumstances
Heider (1958) is usually credited with inventing attribution our’s characteristics.
theory. He argued that people are most concerned with iden- Correspondent inference theory proposes that observers con-
tifying the personal dispositions (enduring characteristics such sider the range of behaviours available at the time of making a
as ability and personality traits) that account for other people’s decision in order to work out the actor’s intention. Each of these
behaviour. In other words, observers want to know what it is behaviours would have brought a number of different effects if
about actors that leads them to act the way they do. Drawing selected. Some of these effects are desirable (your friend’s camera
dispositional inferences carries two basic advantages. First, it has lots of useful features) and some undesirable (the camera cost
a great deal of money). According to Jones and Davis, observers
work out why actions are performed by comparing the effects of
the selected action with those of alternative unselected actions
(taking into account their perceived desirability). In particular,
actors are assumed to have selected their action on the basis of the
PIONEER effects that this action alone produced (effects that would not have
happened if another action had been selected).
Fritz Heider (1896 –1988), the ‘founding father’ of attribu- For example, think back to when you chose to go to the particu-
tion theory, was born in Vienna, Austria. He was invited to lar university or college where you are currently studying instead
the USA in 1930 to join the Gestalt psychologist Koffka’s of a different one. The theory suggests that we could infer your ori-
laboratory at Smith College, then worked at the University ginal intention by comparing the features of these two universities
of Kansas from 1947, where his most influential work on and working out what distinguishes them. For example, the cho-
attribution was conducted. Heider is famous for two theor- sen university might be located in a large city and the other one in
ies in different areas of social psychology: a quieter, more rural setting. If the rejected university also had sev-
attribution theory and ‘balance theory’ (a eral advantages over the one you chose (e.g., a higher reputation,
consistency theory about how relationships a stronger psychology department, better accommodation), then
between more than two people are kept in we might well conclude that living in a city is important enough
equilibrium). He was awarded the American to you to outweigh all these other considerations. More generally,
Psychological Association’s Distinguished correspondent inference theory argues that people try to work out
Scientific Contribution Award in 1965. what it was about a chosen course of action that made it seem
preferable to alternative courses of action. Jones and Davis call this
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 48

48 CHAPTER 3 SOCIAL PERCEPTION AND ATTRIBUTION

Table 3.1 Analysis of non-common effects after observed selection of University X

Features of University X Features of University Y Are features common or Implication about intention
(chosen) (not chosen) non-common?

Comfortable accommodation Comfortable accommodation Common None

Sports facilities Sports facilities Common None

Good reputation Good reputation Common None

Friends applying Friends applying Common None

Urban location Rural location Non-common University X chosen because candidate

wanted to live in an urban location

process the analysis of non- 70

analysis of non-common effects

Essay-writer’s perceived positive attitude towards Castro

observers infer intentions behind actions by common effects (see Table 3.1). Pro-Castro
comparing the consequences of the In reality, our analysis may Anti-Castro
60
behavioural options that were open to the be more complex than implied
actor and identifying distinctive outcomes by this example. For instance,
we might focus on why you 50
chose to go to university at all rather than why you selected this
particular one. How then do observers know what alternatives
40
to compare when trying to explain a course of action? Research
suggests that people are more interested in explaining unusual
than predictable events, and that they explain them by comparing 30
what actually happened with what they think would normally have
happened (Hilton & Slugoski, 1986, and p. 51 below). For example, 20
if a close friend walks past us in the street without saying hello,
we think about what is different this time from the usual times
when she stops to chat. Perhaps she forgot to put on her glasses 10
this morning, or perhaps you have had such a radical change of
hairstyle that you are now almost unrecognizable. 0
Although correspondent inference theory was only intended Choice No choice
to apply when actors are free to choose their behaviour, an experi-
ment conducted by Jones and Harris (1967) casts doubt on this Figure 3.2 Correspondent inferences of essay-writers’ attitudes
assumption. Students at an American university were asked to assess (Jones & Harris, 1967).
another student’s opinion about Fidel Castro’s communist regime
in Cuba after reading a pro-Castro essay that the other student had
supposedly written (see Figure 3.2). One group of participants was
told that the writer had freely chosen what position to adopt in the
essay, whereas another group was told that the essay title had Covariation theory
explicitly requested pro-Castro arguments. According to correspon-
dent inference theory, participants in the latter condition should How do perceivers weigh up different possible causes of behaviour
have ignored the essay’s content when estimating the writer’s atti- and decide on an explanation?
tudes. However, participants tended to conclude that the essay-
writer had pro-Castro attitudes even when the situational constraint Imagine that you know a very conscientious student (Hermione)
was evident. Given that most American students were strongly who always goes through everything on reading lists well in
anti-Castro when the study was conducted, this conclusion seemed advance of classes. Before you had a chance to look at this chapter,
an unlikely one. The investigators concluded that people tend to she already told you what it was about and expressed the opinion
overestimate personal causes that attribution theory was a really boring topic. Clearly, you want
correspondence bias the proposed of behaviour but underesti- to know why she said this. Is it because attribution theory actually
tendency to infer a personal disposition mate situational ones, an im- is tedious? Is it because Hermione is rarely excited by anything?
corresponding to observed behaviour even
when the behaviour was determined by the
portant phenomenon later Or is it because her showy lack of enthusiasm was designed to
situation termed the correspondence impress your jaundiced room-mate (Ron) who happened to be
bias (see p. 55 below). listening in on this particular conversation? Jones and Davis’s
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 49

ATTRIBUTION THEORY 49

model would say that you need to compare the consequences of Why then did Hermione say that attribution theory was
saying that the material was boring with the consequences of saying boring? Kelley argues that you need to consider three kinds of
or doing something else. However, even if we knew about all the evidence, each corresponding to one of the possible causes
relevant alternatives, this would only allow us to narrow down (the object, situation or person). First, you need to know whether
Hermione’s possible intentions, not to say which aspects of the Hermione expresses boredom only about attribution theory (high
event exerted causal influence. Further, even though analysis of distinctiveness) or whether she says that a lot of things are boring
non-common effects might have some use in this particular (low distinctiveness). In other
example, it cannot easily be applied to attributions about non- words, you collect distinctive-
distinctiveness information evidence
deliberate behaviours or feelings. ness information by sampling
relating to how an actor responds to
Kelley’s (1967) covariation across objects. Second, you different entities under similar
theory provides a more gen- need to know whether your circumstances
covariation theory proposes that
observers work out the causes of behaviour eral account of how people friend only says attribution
consistency information evidence
by collecting data about comparison cases. weigh up different possible theory is boring in front of relating to how an actor’s behaviour
Causality is attributed to the person, entity causes of an observed action Ron (low consistency) or towards an entity varies across different
or situation depending on which of these or experience. Its assump- makes similar comments situations
factors covaries with the observed effect
tion is that an actor (e.g., across a range of situations consensus information evidence relating
Hermione) has responded in regardless of who might be to how different actors behave towards the
some way to an object (e.g., attribution theory) in a particular listening (high consistency). same entity
situation (e.g., while Ron was listening). The observer then wants In other words, you collect
to know whether what happened was caused by something about consistency information by
the actor, something about the object, or something about the sampling across situations. Third, you need to know whether it is
situation (or some combination of these three factors). According only Hermione who finds attribution theory boring (low consensus)
to Kelley, observers work this out by systematically collecting and or if other students on your course say the same thing (high con-
processing additional data. The aim is to discover what factors sensus). In other words, you collect consensus information by
need to be in place for the effect to happen. sampling across actors.
As its name suggests, Kelley’s covariation theory argues that Having collected all the relevant data, you are now in a position
observers make their judgements on the basis of covariations or to make your attribution. For example, if Hermione says lots
correlations between effects and their possible causes. In other of things are boring (low distinctiveness), says that attribution
words, ‘the effect is attributed to that condition which is present theory is boring regardless of circumstances (high consistency)
when the effect is present and which is absent when the effect is and none of your other friends says it is boring (low consensus),
absent’ (Kelley, 1967, p. 194). Inferences of causality thus depend you may conclude that it is something about Hermione that
on finding out that the effect’s occurrence relates to the presence makes her bored (a ‘person attribution’). You infer this from
of one or more of the possible causal factors, but not to the pres- the close correlation between the presence of Hermione and
ence of other factors. statements that something is boring (whenever she is included
in a sampled episode, something is described as boring, but when-
ever she is absent, nothing is described as boring). The causal im-
plications of some other possible combinations of consensus,
consistency and distinctiveness (CCD) information are presented
in Table 3.2.
One limitation of the covariation model is that the pattern of
PIONEER information supposed to indicate various attributions is incom-
plete (see Försterling, 2001; Hilton, 1988). For example, knowing
Along with Bernard Weiner, Harold Kelley (1921–2003) that Hermione only says attribution theory is boring in front of
was one of two pioneering attribution theorists working at Ron does not definitively establish the causal role of this situation
the University of California at Los Angeles. His covariation because you have not collected data about how different people
theory of attribution stands as the most influential general react to Ron’s presence. In fact, there is good evidence that people
approach to lay causation, although from the start he can infer many of the predicted implications of other patterns of
acknowledged that it did not apply across all possible situ- CCD information when evidence is provided in this particular
ations. His second, causal schema theory was specifically form (e.g., McArthur, 1972; see Hewstone, 1989, and Kassin, 1979,
intended to explain how people arrive at for reviews).
causal explanations when they are unable Although Kelley’s theory provides a logical basis for attribu-
to carry out the systematic collection of data tion, it is difficult to imagine that people collect evidence so sys-
implied by covariation theory. In addition tematically and engage in such detached processes of analysis every
to these two influential theories, Kelley time they make sense of an event’s causes. That certainly would get
also worked on person perception, attitude boring. Many subsequent developments in attribution theory have
change and relationships. therefore involved correcting this limitation of the covariation
approach.
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 50

50 CHAPTER 3 SOCIAL PERCEPTION AND ATTRIBUTION

Table 3.2 Four patterns of consensus, consistency and distinctiveness information, and their perceived implications (after Kelley, 1967)

Consensus Consistency Distinctiveness Attribution

(across persons) (across situations) (across objects)

Low High Low Person attribution

(No one else says (Hermione says attribution (Hermione says that lots of Effect covaries with person: something
that attribution theory is boring in many things are boring) about Hermione causes her to say that
theory is boring) different contexts) attribution theory is boring

Low Low High Context attribution

(No one else says (Hermione only says (Hermione doesn’t say that Effect covaries with situation: something
that attribution attribution theory is other things are boring) about the presence of Ron causes Hermione
theory is boring) boring in front of Ron) to say that attribution theory is boring

High High High Entity attribution

(Everyone else (Hermione says attribution (Hermione doesn’t say that Effect covaries with object: something
says attribution theory is boring in many other things are boring) about attribution theory makes Hermione
theory is boring) different contexts) say that it is boring

Low High High Person–entity interaction

(No one else says (Hermione says attribution (Hermione doesn’t say that Effect covaries with Hermione together with
that attribution theory is boring in many other things are boring) attribution theory: something about their
theory is boring) different contexts) combination causes her to say attribution
theory is boring

Access to covariation information necessary causes schema).

Knowledge of factors work-
augmentation principle the assumption
that causal factors need to be stronger if an
How do we make causal attributions when information is ing against an effect leads inhibitory influence on an observed effect is
people to conclude that plau- present. The converse of the discounting
incomplete? principle
sible causes must be stronger
One obvious problem for the covariation approach was soon than otherwise (the augment-
recognized by Kelley (1972) himself: often we want to make causal ation principle). For example, if Hermione wanted to impress some-
inferences about events under circumstances when CCD infor- one who was enthusiastic about attribution theory, but still told
mation is either unavailable or them it was boring, then you would probably conclude that she had
causal schema a knowledge structure too time-consuming to collect. a sufficiently strong negative opinion to override her desire to please.
shaping attributions. Causal schemas may On these occasions, Kelley
be either abstract representations of general
argued that we fill in missing
causal principles (e.g., multiple necessary
information by reference to
Knowledge, expectation and
and multiple sufficient causes schemas) or
domain-specific ideas about how particular our existing ideas about how covariation
causes determine particular effects effects are produced (causal
schemas). How do we use our general knowledge to guide our attributions?
For example, when Hermione says that attribution theory
is boring in front of someone else she knows will be impressed by Kelley’s causal schema theory implies that people take shortcuts
such talk, this already tells you about an important factor poten- to inferential conclusions when information or resources are
tially causing this behaviour (i.e., a desire to impress this person). limited (as in other contemporary dual-process models of social
Her statement need not reflect a strong personal dislike for attri- cognition: see Chapter 4). However, Kelley still believed that
bution theory since this situational factor already partly explains observers engaged in more systematic analysis of covariation
what she said. More generally, Kelley argued that observers dis- whenever possible. In fact, there is little evidence that people
count possible causes when they know of other factors working spontaneously collect CCD information even when it is readily
towards an observed effect (the discounting principle) as long as available. Lalljee, Lamb, Furnham, and Jaspars (1984) presented
this effect can be produced by participants with descriptions of events that required explanation
discounting principle the presence of a a range of alternative factors (e.g., ‘John did well on his history essay’). Participants were asked
causal factor working towards an observed (multiple sufficient causes schema). to write down the questions that they wanted to ask in order to
effect implies that other potential factors
are less influential. The converse of the
In other cases, more than one explain these events. Fewer than 20 per cent of their questions
augmenting principle condition must be present for were specifically related to CCD. Instead, most were designed
the effect to occur (multiple to evaluate participants’ specific hypotheses about why the
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 51

ATTRIBUTION THEORY 51

events had occurred (e.g., ‘Did John try especially hard on this things in order to impress others, and that people’s actions don’t
occasion?’). immediately change the weather. However, because much of
On reflection, this is not particularly surprising. A problem with this specific knowledge needs to be learned from observation,
CCD information is that it only tells us whether the actor, object knowledge-based theories of attribution still need to explain how
or situation (or some combination of these) caused the event, but people acquire their knowledge about what typically causes what
not what it is about the actor, object or situation that caused it. As in the first place.
Lalljee and Abelson (1983) point out, knowing that John lied to According to Cheng (1997), covariation information alone
Mary because of something about Mary begs the question of what cannot answer this question because it is insufficient to imply
this something might be that makes people want to lie to her. To causation. Perceivers typically supplement covariation analysis
work this out, we would need to refer to our prior knowledge with their own innate implicit theory that certain events carry
about why people might deceive one another. But then why not unobservable causal powers
just start by consulting this useful knowledge instead of first con- (see also White, 1989). For causal power an intrinsic property of an
ducting a time-consuming covariation analysis? If we need to rely example, the fact that a mag- object or event that enables it to exert
on ready-made explanations anyway, and these can tell us what net consistently attracts or influence on some other object or event
we really want to know, then there is little point in going through is attracted to metal objects
the preliminary step of collecting and sifting through all possible leads us to conclude it has an invisible quality (‘magnetism’)
combinations of CCD information. that brings about these effects. Because our predisposition is to
It is now generally accepted that people don’t usually engage in uncover causal powers rather than to record observable regular-
a thorough data-driven process every time they make an attribu- ities for their own sake, our sampling of covariation data can be
tion. Because we already have expectations that events will unfold more principled and focused. In particular, covariations between
in a certain way, these can be used as a reference point for our competing potential causes and the observed effect are compared
attributions. Indeed, Hilton and Slugoski (1986) argue that people ( probabilistic contrast) in order to determine the nature of the
rarely need to ask themselves the causal question implied by co- underlying causal process. For example, a child might find out
variation theory: ‘why did this happen instead of not happening?’ that audible distress brings about parental attention by repeatedly
(a question that would lead them to weigh up all possible factors crying in similar situations (so that all other plausible causes
that might have led to the event). Instead, they usually want to remain constant) and registering any consistent effects on Mum
know ‘why did this happen instead of what usually happens (under or Dad. The child might also compare this strategy with throw-
these circumstances)?’ Thus, people look for causes among the dif- ing toys around. Thus, even small infants may conduct informal
ferences between actual and anticipated event sequences (abnor- experiments based on an innate theory that effects are caused
mal condition focus) rather than exhaustively sifting through all by events with intrinsic causal powers. Cheng argues that the
available evidence. Observers know where to look for relevant more specific causal knowledge guiding our subsequent attribu-
causes not only because they understand general principles of tions is originally acquired by making probabilistic contrasts of
causality (as implied by Kelley’s causal schema model), but also this kind.
because they have access to cognitive scripts telling them how par-
ticular kinds of event (e.g., conversations, parties, restaurant visits)
ordinarily unfold in the social world (e.g., Cheng & Novick, 1990;
Read, 1987).
Attributions for success and
failure
Covariation and causal power What are the implications of attributing success and failure in
different ways?
How do we use more specific causal knowledge to guide our causal
explanations? Some of the events that we are most motivated to explain are suc-
cesses and failures. For example, if you get a better than usual
A final limitation of covariation theory is captured by a slogan grade in an exam, you may wonder whether this was due to your
familiar from statistics classes: ‘correlation is not causation’. particular affinity for the topics covered, your thorough exam
Establishing that factor X covaries with effect Y can never prove revision or the fact that exactly the right questions happened to
that X caused Y, because a correlated third variable may have come up. Your conclusion will help you work out how likely it is
exerted the real influence (or indeed Y might have caused X). For that you will be able to maintain this level of performance and how
example, a covariation between revision and fine weather does you might go about achieving this.
not mean that studying hard can make the sun shine. The most influential theory of achievement-related attribution
Again, prior knowledge can help us untangle causal relations was developed by Weiner (1979, 1985), who argued that our con-
of this kind. Because we are already aware of what kinds of factor clusions about the causes of success and failure directly affect
are possible causes of particular effects, we can reject certain factors future expectations, motivations and emotions. One of Weiner’s
as irrelevant and focus down our causal search. We know, for main contributions was his classification of the perceived causes
instance, that energy is required to induce movement, pressure of success and failure (see Table 3.3). According to this classifica-
to produce deformation of objects, that people sometimes say tion, perceived causal factors may be: (1) internal or external
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 52

52 CHAPTER 3 SOCIAL PERCEPTION AND ATTRIBUTION

(locus); (2) stable or variable (stability); and (3) controllable or un- the outcome, whereas attributing performance to an external fac-
controllable (controllability). tor means that something to do with the situation was responsible.
Attributing your exam success to an internal factor means that Both internal and external factors can be either variable or stable.
you believe that something relating to you as a person determined For example, attributing your performance to intelligence means
that you think something internal and relatively unchanging about
you led you to do well.
Weiner also argued that causal factors are perceived as either
controllable or uncontrollable, and that this distinction too makes
a difference to your reaction to achievement outcomes. For ex-
ample, if you believe that your exam success was due to an inter-
nal, stable and uncontrollable factor (your innate aptitude for this
kind of material), then you may feel that there is no need to try
hard in order to repeat your success. On the other hand, if you
think that your good grade was due to an internal, variable and
controllable factor (e.g., effort), you will probably conclude that
you need to stay motivated in order to succeed in future. Thus,
attributions about success and failure are not simply intellectual
conclusions about performance, they also make a real difference to
our expectations and motivation.

Plate 3.2 Do our conclusions about the causes of success and

failure, e.g. in an exam, directly affect future expectations, Attributional reformulation
motivations and emotions?
of learned helplessness theory
Are certain patterns of attribution symptoms, or causes,
of depression?
PIONEER Weiner’s conclusions about the motivational consequences of
attributions for success and failure have broader implications for
Bernard Weiner (b. 1935) is currently Professor of Psycho- understanding clinical disorders. One influential application has
logy at the University of California, Los Angeles, where he been the attributional reformulation of learned helplessness theory
has worked since 1965. He received his undergraduate of depression. Learned help-
degree from the University of Chicago and his doctorate lessness theory (Seligman,
from the University of Michigan in 1963. In 1965, following 1975) originally argued that learned helplessness theory the proposal
two years at the University of Minnesota, he went to UCLA. that depression results from learning that
depression results from learn-
outcomes are not contingent on one’s
He is most famous for his influential research ing that nothing you do makes behaviour
into attributions for success and failure, iden- any difference to outcomes.
tifying the main types of attribution, their The idea was that if rewards
underlying dimensions and their effects on and punishments have no relation to your actions, you soon
motivation and emotion. His classification of learn to give up trying to attain the former and avoid the latter.
perceived causes of behaviour has had an However, there are many uncontrollable situations in everyday
even greater impact. life that don’t make people depressed. For example, many people
enjoy betting on games of chance, where the outcomes are

Table 3.3 Possible causes of success and failure (after Weiner, 1979, 1985)

Internal External

Stable Unstable Stable Unstable

Controllable Mastery (e.g., knowledge, Effort Enduring situational Temporarily available

skill) and social resources situational and social resources
(e.g., contacts, wealth) (e.g., advice, assistance)

Uncontrollable Aptitude (e.g., Energy Task ease or difficulty Luck/chance

intelligence, coordination)
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 53

ATTRIBUTION THEORY 53

Objective Perceived Attribution Expectations of Helplessness

non-contingency non-contingency Non-contingency non-contingency symptoms
Nothing the person The person notices is attributed to The person concludes that Depression coupled
does makes a difference that nothing she does internal, global and future actions will make no with motivational
to what happens. makes a difference. stable factors. difference to what happens. deficits.

Figure 3.3 Five steps to depression: the attributional reformulation of learned helplessness theory (Abramson et al., 1978).

completely beyond their influence. Gambling does not always Consider, however, the contrasting attribution to internal, stable
make people depressed even if money is lost. and global causes: the man finds her unattractive not out of any
This observation suggests that helplessness alone does not auto- passing whim but because of the kind of person she is. His dislike
matically lead to depression; other factors must also be present. is permanent and applies across all situations. Because being dis-
One clue to what these other factors might be comes from another liked is seen as reflecting something about her, other men will
key clinical feature of depression that learned helplessness theory probably dislike her too and her future chances of romantic hap-
cannot explain, namely an exaggerated sense of personal respon- piness are slim indeed. Further, because the factors are global they
sibility for negative outcomes. If uncontrollable events cause de- apply not only to romantic attraction but to other areas of her life
pression, why should depressed people think that they have caused as well. She can only look forward to consistent and universal bad
these events to happen? Abramson, Seligman and Teasdale’s (1978) outcomes which she can do nothing about. These negative ex-
answer is that helplessness only makes people feel chronically pectations unsurprisingly lead to depression.
depressed if it is attributed to intrinsic features of the self. In other The theory thus argues that people who have developed a
words, a specific pattern of attribution for uncontrollability may tendency to attribute uncontrollable events to internal, stable and
determine clinical depression (see Figure 3.3). global attributions have a greater risk of subsequently developing
In defining this pattern, Abramson and colleagues extended chronic depression. However, there is little evidence that attribu-
Weiner’s classification to include another distinction between tions made prior to the onset of depression are distorted in this
specific and global causes. Global causes apply to a wide variety of way (Lewinsohn, Steinmetz, Larson & Franklin, 1981), so it seems
situations, whereas specific factors relate only to the particular situ- equally plausible that self-focused explanations are symptoms
ation at hand. The quality and persistence of depression depends rather than causes of depression (but see Rude, Valdez, Odom &
on whether the cause of uncontrollability is perceived as internal Ebrahimi, 2003).
or external, stable or variable, and global or specific. Although depressed and non-depressed people evidently inter-
To illustrate this classification, Abramson and colleagues used pret negative events in different ways, who is more accurate? Some
the example of a woman who has been rejected by a man in whom theorists have argued that it is not depressed people who are
she is romantically interested. Nothing she does makes any differ- unduly pessimistic and unable to see the glass as half-full rather
ence to the way he feels about her. According to the model, the than half-empty, but rather non-depressed people who protect
way this woman reacts to this experience of helplessness depends themselves from unpleasant realities by seeing everything in an
on what she thinks the causes of rejection might be (see Table 3.4). unrealistically positive light (the ‘illusory glow’, Taylor & Brown,
The least-threatening interpretation would be that her rejec- 1988). According to this view, termed depressive realism, de-
tion was caused by something external, unstable and specific. This pressed people are ‘sadder but
man in particular was not attracted to her at this moment and in wiser’. In support of this idea,
this particular situation (e.g., he may simply not have been in the Lewinsohn and colleagues depressive realism the idea that
depressed people’s interpretations of reality
mood for romance at the time). The consequences of this conclu- (1980) found that depressed are more accurate than those of non-
sion are not too serious for the woman, because she is still able to participants’ ratings of their depressed people
anticipate greater success with this man or other men in future. social functioning during a

Table 3.4 Possible causes of romantic rejection (from Abramson et al., 1978)

Internal External

Stable Unstable Stable Unstable

Global I’m unattractive to My conversation Men are overly competitive Men get into rejecting
men sometimes bores men with intelligent women moods

Specific I’m unattractive to him My conversation He’s overly competitive He was in a rejecting mood
sometimes bores him with intelligent women
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 54

54 CHAPTER 3 SOCIAL PERCEPTION AND ATTRIBUTION

group discussion were closer to those of observers than were non-

depressed participants’ ratings. Although observers rated the per- PIONEER
formance of non-depressed participants more positively than that
of depressed participants, they did not rate it as positively as the
non-depressed participants themselves did. Stanley Schachter (1922–1997) is best known for his two-
factor theory which inspired the cognitive approach to emo-
However, Campbell and Fehr (1990) found that participants
tion, and for his clever experiment with Jerome Singer which
with low self-esteem (a typical feature of depression) were more
remains a classic despite the apparent inconclusiveness
accurate only when their judgements were compared against those
of its results. His earlier work on affiliation also provided a
of an observer who did not participate in the interaction. Indeed,
forerunner of contemporary interpersonal approaches to
evidence suggests that outside observers may be unduly harsh emotion, showing that people seek out other people in order
judges because they think that their task is to be critical. When par- to make sense of their own feelings. Throughout his career,
ticipants’ own ratings were compared with those of the person Schachter pursued the important idea that external cues
they were having the conversation with, participants with high (including social cues) can shape the inter-
rather than low self-esteem came out as more accurate. It seems pretation of supposedly internal states such
then that depressives’ judgements are probably only more accur- as emotions and feelings of pain and hunger.
ate when circumstances match their negative outlook. How- Such an approach provides a valuable coun-
ever, it is also worth bearing in mind that depressives’ negative terargument to the more usual assumption
judgements can easily turn into self-fulfilling prophecies. If some- that emotion, sensation and motivation are
one doesn’t even try because they are sure they will fail regardless primarily biological and individual processes.
of effort, this makes it more likely that they will in fact fail.
Whether the attributional pattern associated with depression
is a symptom or cause of depression, and whether it is realistic or
unrealistic, reformulated learned helplessness theory suggests that
therapy should focus on changing it in order to alleviate the symp- Physiological (ANS) arousal
toms of depression. In practice, current cognitive and psycho- Intensity
dynamic therapies for depression attempt to modify a wide range
of negative interpretations, but correcting maladaptive attributions
for failure may explain part of their apparent effectiveness (e.g., Emotional state
Barber et al., 2005).
Quality
Cognitions about the situation
Misattribution of arousal
Figure 3.4 Schachter’s two-factor theory of emotion.
How do we ascertain what we are feeling and why?

The attributional reformulation of learned helplessness theory

suggests that we feel more depressed about unpleasant events if Schachter (1964) therefore argued that emotions depend upon
we conclude that their causes are internal, stable and global. But the attributions we make for our internal feelings, rather than
how do we recognize our reaction to this attributional pattern directly reflecting these feelings themselves. Thus, perceptions of
as depression rather than something else? This may sound a silly arousal (the physiological factor) tell us that we may be experi-
question because it usually seems that the nature of our current encing an emotion, but not what emotion it is. We therefore try to
emotional state is self-evident. By contrast, misattribution theor- work out why our bodies are aroused (the cognitive factor) in
ies imply that we sometimes need to work out what it is we are order to answer this second question (see Figure 3.4). If we con-
feeling and that this inference process is susceptible to social clude that our arousal is caused by the attractive person we are
influence. having coffee with, we may interpret it as a symptom of love (or
One of the earliest psychological theories of emotion was at least lust: see Chapter 10, this volume). However, if we think
devised by William James (1884). He argued that each emotion our arousal is due to the fact that someone else keeps butting into
has its own distinctive profile of bodily changes and that we can our private conversation, we may conclude that it reflects anger.
directly sense our emotion by registering these changes. However, Finally, if we attribute our symptoms to the caffeine in the strong
Cannon (1927) pointed out that the patterns of internal physio- cup of coffee we are drinking, we may conclude that our arousal
logical activity associated with very different emotions are actu- is non-emotional.
ally rather similar. For example, both fear and anger involve In a famous experiment, Schachter and Singer (1962) tried
increases in heart rate, blood pressure and other kinds of metabolic to determine whether an identical physiological state could be
activity (physiological arousal in the autonomic nervous system, perceived as anger, euphoria or non-emotionally depending on
ANS) designed to release energy to the muscles in preparation for participants’ interpretations of its causes. Autonomic arousal was
vigorous activity. This means that we cannot tell these emotions manipulated by administering an adrenaline injection to one group
apart simply by checking what is happening inside our bodies. of participants and a placebo injection to others under the guise
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 55

ATTRIBUTION THEORY 55

that the syringe contained a new vitamin compound (Suproxin) that allowing clinicians to minimize otherwise maladaptive reactions
the experimenters were testing. (e.g., Ross, Rodin & Zimbardo, 1969). For example, Storms and
Adrenaline-injected participants in one group were told that Nisbett (1970) reported that students with mild insomnia fell asleep
Suproxin might lead them to experience side effects such as a more quickly after being told that they had taken an arousal-
pounding heart and shaking hands (genuine arousal symptoms). inducing pill (reverse placebo effect). The investigators argued that
These participants were therefore able to interpret their bodily misattribution of arousal symptoms to the pill neutralized the
symptoms correctly as non-emotional responses to the injection. mild-insomniac participants’ usual interpretation in terms of
However, participants in another group were given incorrect anxiety. However, Calvert-Boyanowsky and Leventhal (1975)
information about the adrenaline injection’s effects (i.e., that demonstrated that such effects may be explained by the correct
there would be no side effects or arousal-irrelevant side effects). anticipations set up by symptom warnings rather than misattribu-
Participants in this condition should therefore experience arousal tion per se. For example, knowing what is about to happen to your
symptoms without knowing their cause, and consequently seek body means that symptoms are less surprising and less emotionally
an emotional explanation. upsetting when they arrive. However, it is less clear whether this
Schachter and Singer stage-managed the situation to encour- explanation can explain the reduced happiness of participants in
age specific attributions for any unexplained arousal. Each participant Schachter and Singer’s informed euphoria condition.
was left in a waiting room with an accomplice of the experimenter In sum, Schachter and Singer’s clever experiment does not
posing as another participant who behaved in one of two ways. In offer conclusive support for all aspects of two-factor theory. This
one condition, the accomplice improvised a basketball game using may be partly because it is difficult to manipulate arousal and emo-
scrap paper and a wastebasket, and encouraged the other participant tional cognitions independently when the two usually go hand in
to join in. In the other condition, the accomplice became progres- hand. Subsequent studies have been similarly inconclusive (e.g.,
sively more irate while working through an increasingly insulting Erdmann & Janke, 1978; Marshall & Zimbardo, 1979; Maslach,
questionnaire that the participant also had to complete. Its final 1979), and many theorists now believe that Schachter overstated
item read: ‘With how many men (other than your father) has your how easy it was to influence emotional interpretations. Because
mother had extra-marital relationships?’ The only response altern- our attributions about, and appraisals of, emotional situations usu-
atives provided were: ‘10 and over’, ‘5–9’ and ‘4 and under’. ally determine our autonomic as well as emotional reactions in the
According to Schachter’s theory, emotion should only occur first place (e.g., Lazarus, 1991), we often know in advance what
when autonomic arousal is attributed to an emotional cause. In we are going to feel.
other words, emotion should not have been experienced by either
placebo-injected participants (because they were not aroused) or
participants who had been correctly informed about the adrena- Attributional bias
line injection’s effects (because they did not attribute their arousal
to the emotional situation). However, adrenaline-injected par- What are the main types of attributional bias, and how can they be
ticipants who were unaware that their symptoms were caused by explained?
this injection should have explained their arousal in terms of the
plausibly euphoric situation when with the playful confederate, Covariation theory and the correspondent inference model both
but in terms of the plausibly irritating situation when completing tended to view attribution as a data-driven process wherein all
the insulting questionnaire. These two groups, therefore, should potentially relevant information is systematically processed. How-
have experienced widely divergent emotional reactions of euphoria ever, as we have seen, subsequent research suggests that causal
and anger, respectively. inferences are shaped by prior knowledge and expectations (e.g.,
In fact, results were less clear-cut (see Reisenzein, 1983). For Hilton & Slugoski, 1986; Read, 1987), or by learned attributional
example, placebo-injected participants did not report significantly styles (e.g., Abramson et al., 1978), and that they may be extrane-
less emotion than participants who were injected with adrenaline ously influenced by contextual variables (e.g., Schachter & Singer,
but not informed about the injection’s genuine side effects. Further, 1962). Thus, people seem to attach more weight to some causes at
emotion reports of misinformed adrenaline-injected participants the expense of others when drawing causal conclusions. Precisely
did not differ substantially between euphoria and anger conditions what kinds of causes are typi-
(participants reported themselves to be mildly happy in both con- cally favoured under different
attributional bias systematic distortions in
ditions; see Zimbardo, Ebbesen & Maslach, 1977). circumstances has been the the sampling or processing of information
One significant result obtained by Schachter and Singer clearly focus of research into various about the causes of behaviour
did accord with predictions, however. Participants injected with attributional biases.
adrenaline and correctly warned of the effects consistently reported
less positive emotion in the euphoria condition, and less negative The correspondence bias In their professional lives, psycho-
emotion in the anger condition than participants misled about side logists of different persuasions sometimes disagree about whether
effects. Schachter’s explanation was that the informed group cor- internal or external explanations of human behaviour deserve
rectly attributed their arousal to the injection and labelled it in non- more emphasis. For example, most experimental social psycho-
emotional terms. logists focus on situational influences and often ignore people’s
Subsequent experiments have suggested that genuinely emo- characteristic dispositions. By contrast, personality psychologists
tional arousal may also be misattributed to non-emotional sources, attach more weight to personal traits, usually without giving much
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 56

56 CHAPTER 3 SOCIAL PERCEPTION AND ATTRIBUTION

RESEARCH CLOSE-UP 3.1

The correspondence bias in attributing knowledge participants from the experimental condition were individu-
to the quiz master or the contestant ally simulated by confederates posing as participants. Each
simulated quiz was watched by a pair of observer participants
Ross, L.D., Amabile, T.M. & Steinmetz, J.L. (1977). Social roles,
who did not know that it was a simulation. As soon as the quiz
social control, and biases in social-perception processes. Journal
was over, questioners, contestants and observers all separately
of Personality and Social Psychology, 35, 483–494.
rated the general knowledge of the questioner and contestant
compared to the average student at the same university, using
a 100-point scale.
Introduction
People’s social roles shape the way that they interact with one
another. These influences are most obvious in situations when
Results
one person has relatively greater control over an interaction. An On average, contestants got only 4 out of 10 questions right in
oral examination, for example, permits examiners to decide the quiz. General knowledge ratings of questioners and contes-
what topics should be discussed while the examinee has rela- tants were analysed. Contestants in the experimental condition
tively less influence. One consequence is that examiners have rated their own general knowledge as significantly worse than
greater opportunity to display their specialized knowledge. In that of questioners, and observer participants also rated con-
accordance with the correspondence bias, Ross and colleagues testants’ general knowledge as significantly inferior. However,
argue that people take insufficient account of these role- questioners did not rate their own general knowledge as higher
conferred advantages when arriving at attributions for behaviour. than that of contestants (see Figure 3.5 for the mean ratings
As a consequence, people with relatively greater social control for these conditions). Further, control participants showed
appear wiser and more able than they really are (and their significantly smaller differences between their ratings of ques-
social position therefore seems more justified). To test this tioners and participants than did experimental participants.
hypothesis, the investigators simulated a quiz game in which Although no integrated analysis of the results was presented by
participants were allocated the role of either questioner or con- the investigators, the pattern of findings clearly accords with
testant. Questioners were given the opportunity to devise their predictions.
own questions, thereby permitting an unrepresentative demon-
stration of their idiosyncratic expertise. The prediction was that
questioners would be viewed as higher in general knowledge.
Discussion
The findings demonstrate that the situational advantage con-
Method ferred by being allowed to devise your own questions led to
higher general knowledge ratings from both contestants and
Participants observers (who probably attempted to answer the questions
Eighteen pairs of male students and 18 pairs of female students
from an introductory psychology class were recruited for a study
into processes whereby ‘people form impressions about gen- 100
eral knowledge’. Twelve pairs of participants of each gender 90
compared to ‘average’ student (50)

Questioner
were assigned to the experimental condition and six pairs of 80 Contestant
Rated general knowledge

each gender were assigned to the control condition. For the

70
observer condition, another 24 pairs of participants subse-
quently watched individual simulations of the quizzes originally 60
conducted by female pairs. 50
40
Design and procedure
In the experimental condition, the role of questioner or con- 30
testant was allocated to one person in each pair using an 20
explicitly random procedure. Questioners were told to devise 10 10
‘challenging but not impossible’ general knowledge questions,
0
which they then asked contestants in a quiz. For example, Questioners’ Contestants’ Observers’
one question was: ‘What is the longest glacier in the world?’ ratings ratings ratings
In the control condition, contestants were asked questions
devised by earlier participants from the experimental condition. Figure 3.5 Mean ratings of general knowledge after quiz
In the observer condition, the quizzes conducted by female game (adapted from Ross et al., 1977).
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 57

ATTRIBUTION THEORY 57

privately to themselves too). This seems to provide a clear their questions meant that they must be inferior in general
example of the correspondence bias. Because participants’ roles knowledge, presumably because their own role-conferred
were allocated randomly in this study, it is statistically unlikely advantage was extremely salient to them. Thus questioners
that one group (the questioners) should happen to be genuinely apparently recognized their own relatively advantaged position,
higher in general knowledge. Indeed, Ross and colleagues and were able to correct any attributional bias.
administered brief tests of general knowledge to all participants Subsequent studies have shown that observers are aware
after the quiz and found no differences in performance. of limitations to the apparent superiority of the questioners
The study also carries implications about the limits of (Johnson, Jemmott & Pettigrew, 1984; Sumpton & Gregson, 1981),
the correspondence bias. Questioners in the experimental con- and that bias in this setting depends partly on what questions
dition did not conclude that contestants’ inability to answer are asked (e.g., Schwarz, 1994).

attention to the impact of the environment (see Chapter 1, this it happens regardless of current circumstances. However, other
volume). The correspondence bias suggests that the naïve psy- demands on cognitive resources may interfere with the situational
chology practised by laypeople is closer to personality psychology correction process, leading us to underestimate the power of ex-
than to experimental social psychology. Behaviour is often seen as ternal factors. (The different stages at which these sources of bias
a reflection of an actor’s corresponding internal disposition (e.g., may intrude are shown in Figure 3.6.)
aggressive behaviour reflects aggressive personality) even when it An experiment conducted by Gilbert, Pelham and Krull (1988)
was actually caused by situational factors (e.g., severe provoca- provides support for this last explanation. Participants observed a
tion). Research close-up 3.1 presents a famous example of this ef- silent videotape of a woman talking nervously to a stranger and
fect (see also Jones & Harris, 1967, described earlier). then rated how anxious she was as a person. Subtitles indicating
Why do people underestimate situational influences? Accord- current conversation topics informed some participants that the
ing to Gilbert and Malone (1995), a number of different processes woman was discussing her sexual fantasies (offering a situational
may be involved. First, some situational forces are subtle and explanation for her nervousness) but told others that she was
difficult to detect. If observers are not aware of these influences in talking about gardening. Further, some participants were told to
the first place, they can hardly be expected to factor them into their memorize the subtitles, imposing an additional cognitive demand
explanations. Second, our expectations about how other people that should interfere with any situational correction process. As
will behave may distort our interpretations. For example, we may predicted, participants under higher cognitive demand tended to
mistakenly assume that the prospect of public speaking terrifies believe that the woman had an anxious personality regardless of
others just as much as it terrifies us (an example of the false conversation topic, whereas low-demand participants rated her as
consensus bias). Therefore, when someone appears calm before less dispositionally anxious when they believed she was discussing
their turn to speak, we may sex rather than gardening. Presumably the low-demand particip-
false consensus bias the assumption that conclude that their confident ants had sufficient cognitive resources remaining to correct for
other people generally share one’s own personality must be over- their initial automatic dispositional inference.
personal attitudes and opinions riding an otherwise anxiety- Gilbert and colleagues’ theory suggests that attribution always
provoking situation. involves automatic processes but only sometimes involves con-
Finally, Gilbert and Malone suggest that people sometimes fail trolled processes as well (e.g., Shiffrin & Schneider, 1977). It is
to correct their initial inferences about the causes of behaviour, therefore another example of the dual-process models that are
especially when processing demands are high. The idea here is that currently popular in research on social perception and cognition
people’s automatic reaction to observed behaviour is to conclude (see Chapter 4, this volume). The argument that we spontan-
that it reflects an actor’s disposition. Any relevant situational eously and automatically make inferences about people’s traits
influences are then factored in using a more deliberate reasoning is supported by research conducted by Smith and Miller (1983).
process. Because the initial dispositional inference is effortless, In two studies, these investigators demonstrated that participants

Sources Not noticing Expectations about Processing

of bias subtle influences behaviour demands

Initial
Perception of Dispositional Situational
Stages: perception
behaviour inference correction
of situation

Figure 3.6 Processes leading to correspondence biases (adapted from Gilbert & Malone, 1985).
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 58

58 CHAPTER 3 SOCIAL PERCEPTION AND ATTRIBUTION

presented with sentences describing an actor’s behaviour (e.g., 0.45

‘Ted breaks a friend’s expensive camera that he borrowed’) made US participants
0.4
judgements about the actor’s traits more quickly than they made Hindu participants

Proportion of references to general

inferences about the specific causes of the action. Indeed, it may be 0.35 from India
that we only go to the effort of engaging in a deliberate causal ana-

internal dispositions
lysis and check the validity of our automatic trait attributions when 0.3
we are specifically motivated to think about why a particular beha- 0.25
viour occurred (e.g., when the behaviour affects us negatively,
or when it is unexpected), and when we have sufficient cognitive 0.2
resources to engage in the necessary controlled processing. 0.15

Variability in correspondence biases The correspondence 0.1

bias was once believed to be so pervasive and inescapable that
0.05
it was dubbed ‘the fundamental attribution error’ (Ross, 1977).
However, subsequent research suggests that it is more context- 0
dependent than such a description implies (see Gawronski, 2005, for 8 11 15 Adult
a review). For example, a study by Krull (1993) showed that asking Age group
people to diagnose the situation rather than the person led them
Figure 3.7 Dispositional attributions in two cultural settings
to make automatic situational rather than dispositional inferences.
across four age groups (Miller, 1984).
Participants were again exposed to a silent videotape showing a
woman talking, and were told that she was discussing sensitive
topics with her therapist. Those whose task was to assess how
anxiety-provoking the conversation was rated the woman as less dis-
positionally anxious, but the situation as more anxiety-provoking,
when cognitive load was high than when it was low. By contrast,
those whose task was to assess how dispositionally anxious the
woman was rated the woman as more dispositionally anxious, but
the situation as less anxiety-provoking, when cognitive load was
high. It therefore seems that automatic dispositional inferences
only occur if the inferential goal is to understand the person rather
than the situation that person is in.
Many attribution studies have implicitly encouraged such
inferential goals by orienting participants to actors rather than
circumstances. For example, like all other sentences presented in
Smith and Miller’s (1983) study, ‘Ted breaks a friend’s expensive
camera that he borrowed’ begins with, and uses as subject of the
sentence, the name of the actor performing the behaviour. Perhaps Plate 3.3 Members of collectivistic cultures are more sensitive
such sentences convey trait information more directly than they than members of individualistic cultures to the power of situations.
convey situational information (see also Brown & Fish, 1983, dis-
cussed below).
Operation of the correspondence bias also varies across cul- children are socialized to see themselves more as part of groups
tures. For example, Miller (1984) compared explanations offered that must work together to attain valued goals (collectivistic cul-
for deviant and prosocial behaviours by children (aged 8, 11 and 15) ture). This cultural emphasis is likely to lead to more frequent
and adults from the USA and Southern India. She found that applicability of inferential goals directed at situations (especially
US adults attributed events to dispositional causes significantly social situations) rather than individual actors.
more than Indian adults or children from either country, suggest- Despite their relatively higher preference for situational expla-
ing that North Americans but not Indian Hindus learn over the nations, even collectivists frequently assume that actors have dis-
course of development to favour dispositional explanations (see positions corresponding to their behaviour. For example, Korean
Figure 3.7). participants (like US participants) assumed that a controversial
Why should members of some societies develop a stronger essay reflected the writer’s opinions even when they believed
preference for dispositional explanations? Many western societies that the writer had been told what point of view to defend (Choi
such as the USA and many European countries are said to be char- & Nisbett, 1998). However, when the situational constraint was
acterized by a culture of individualism in which personal effort and made more salient, Korean
ability combine to produce deserved outcomes. People socialized participants were better able
salience a property of stimuli in relation to
into such cultures may learn to adopt the inferential goal of to take its influence into perceivers that causes them to attract
understanding actors rather than their circumstances in most account. Choi and colleagues attention
contexts. However, in some other societies (e.g., India, Japan), (1999) argue that members of
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 59

ATTRIBUTION THEORY 59

collectivistic cultures are more sensitive to the power of situations

than are members of individualistic cultures, enabling them to cor-
rect their initial dispositional inferences under some circumstances.

The actor–observer difference The actor-observer difference

compares attributions people make about others with those that
they make about themselves.
actor–observer difference general
Although westerners’ default
tendency for people to explain their own assumption is often that other
behaviour in more situational terms but people’s behaviour reflects a
other people’s behaviour in more corresponding disposition, it
dispositional terms
seems that we tend to em-
phasize external, situational
factors when explaining our own behaviour. This difference was
first identified by Jones and Nisbett (1972) and has received
qualified support since then (e.g., see Watson, 1982). Plate 3.4 Your behaviour, e.g. being lively at a party, may appear
Why don’t explanations of our own conduct follow identical more consistent to others than it really is.
principles to our explanation of other people’s conduct? Two main
explanations have been proposed, and both probably play some how you have acted in a restricted set of contexts. Thus, your
role in accounting for actor–observer differences. First, actors behaviour may appear more consistent to them than it really is.
have access to a wider range of information about the factors The second factor contributing to actor–observer differences
leading to their own actions ( Jones & Nisbett, 1972). For exam- concerns direction of attention. When observing someone else’s
ple, when explaining your liveliness at a party, you are able to con- behaviour, we tend to focus on that person rather than their situ-
sider other situations in which you have acted in a less extroverted ation. Conversely, when we ourselves are acting, our attention tends
manner (e.g., when meeting someone for the first time, or being to be directed outwards. Perhaps then we simply assume that what-
interviewed), and may therefore conclude that you are not a ever is occupying our attention is exerting the most causal influence
consistently lively person. By contrast, most other people only know (e.g., Taylor & Fiske, 1978; see Research close-up 3.2 on Storms, 1973).

RESEARCH CLOSE-UP 3.2

Reversing the actor–observer effect by Method

manipulating perspective Participants
Storms, M.D. (1973). Videotape and the attribution process: Thirty groups of four male students took part in this study. Two
Reversing actors’ and observers’ points of view. Journal of members of each group were randomly assigned the role of
Personality and Social Psychology, 27, 165–175. observer and the other two were assigned the role of actor.

Design and procedure

Introduction Stage 1. Actors were told to have a conversation to get to know
Storms (1973) proposed that differences between actors’ and each other, while facing each other across a table. Each observer
observers’ attribution depend partly on their different physical was seated next to one of the actors and told to observe
points of view: actors’ attention is typically directed outwards the actor across the table from him. Two video cameras were
towards the situation (including other actors), whereas observers’ also set up, each trained on one of the actors (see Figure 3.8,
attention focuses on the observed person (i.e., the actor). Stage 1).
Indeed, one explanation for the correspondence bias is that
actors are often the most dynamic and interesting objects in the Stage 2. Participants in the video condition were told that they
environment and therefore attract observers’ attention (and would now see the videotape of the interaction played back, but
deflect it from other aspects of the situation; see Heider, 1958; because only one camera had worked they would only see the
Taylor & Fiske, 1978). The increasing availability of video tech- tape of one of the actors. Thus, one actor and one observer from
nology in the early 1970s allowed Storms to manipulate actors’ each group saw a replay of the conversation from the same per-
and observers’ perspectives in order to assess the influence of spective as before, while the other actor and observer saw a
this factor on situational and dispositional attributions. video replay from the reversed perspective (i.e., the actor now
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 60

60 CHAPTER 3 SOCIAL PERCEPTION AND ATTRIBUTION

Stage 1 Bird’s-eye view of get acquainted session (arrows indicate 8

direction of attention)
Actors
7 Observers

Dispositional – situational causation

Observer of B Observer of A
6

4
Actor A Actor B
3

1
Camera on B Camera on A
0
No video Same Reversed
perspective perspective

Stage 2 Bird’s-eye view of video replay

Figure 3.9 Reversing the actor–observer difference following
video replay (adapted from Storms, 1973).

Observer of B by personal characteristics and by characteristics of the situation.

Observers rated their respective actors from Stage 1.

Results
Storms calculated difference scores by subtracting summed rat-
Actor A ings of situational attribution for the four key behaviours from
TV monitor summed ratings of dispositional attribution. These difference
showing scores were then analysed. In the no-video and same-perspective
Actor A conditions, actors’ attributions were less dispositional (more
situational) than observers’ (see Figure 3.9). But in the reversed-
perspective condition, observers’ attributions were less dis-
Actor B
positional (more situational) than actors’.

Discussion
This study demonstrates that actor–observer differences can be
Observer of A reversed by showing actors their own behaviour and showing
Same perspective observers the situation that actors are responding to (in this
case, the other actor). A more general conclusion may also be
Changed perspective
possible: that actors and observers tend to attribute greater
causality wherever they pay attention. Indeed, later studies
Figure 3.8 The two stages of Storms’s (1973) procedure.
(e.g., Taylor & Fiske, 1978) have shown that salient (attention-
grabbing) factors tend to be seen as exerting more causal influ-
saw his own face, and the observer saw the face of the actor that ence than non-salient factors.
he had not originally observed; see Figure 3.8, Stage 2). In the One criticism of this study is that the usual actor–observer
no-video condition, participants were told that none of the difference was not demonstrated (e.g., Gilbert & Malone, 1995).
video equipment had worked and that the planned video replay For example, analysis of direct ratings rather than difference
would therefore not take place. scores shows that actors were rated in equally dispositional
terms by themselves and their observers across all conditions.
Measures However, the reported effect on situational attribution is
After Stage 2, actors rated their own friendliness, nervousness, theoretically interesting even if dispositional attribution is
talkativeness and dominance during the conversation, then rated unaffected. The general implication is that we can correct for
the extent to which each of these behaviours had been caused inattention to situational factors by manipulating attention.
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 61

ATTRIBUTION THEORY 61

Of course, direction of attention can only explain preferences

for dispositional explanations of other people’s behaviour when
the other person is physically present. However, observers also
tend to say that the actor was the cause of behaviour when it is
described to them in words. For example, when told that ‘John went
to the cinema’, most people will tend to think that this reflects
something about John rather than something about the cinema.
According to Brown and Fish (1983), the reason is that the English
language implies that subjects of action verbs are responsible for
the action described. By contrast, the objects of experience verbs are
usually seen as causal (e.g., we tend to interpret the sentence John
liked Angela as John’s liking for Angela being caused by something
about Angela).

Self-serving attributional biases What pushes or pulls our

attributions in particular directions? The examples presented so far Plate 3.5 Teacher’s explanations of pupils’ success and failure can
suggest either that we are drawn towards salient factors or that we show self-serving bias.
are following generally valid rules of explanation (e.g., persons
cause effects) in circumstances where they happen to be less
appropriate (see Nisbett & Ross, 1980). The basic idea is that people’s
attributions are occasionally imperfect approximations of the
causal structure of reality, but at least aim to represent that reality B’s performance improved, the students explained this improve-
accurately. However, some kinds of bias are harder to explain in ment in terms of their own abilities as a teacher. But when pupil
these terms. Self-serving attributional biases are those that seem to B failed to improve, they attributed this to his lack of ability rather
represent a motivated distor- than their ineffective teaching methods.
self-serving biases motivated distortions tion of what has happened in Zuckerman (1979) reviewed a number of apparent demon-
of attributional conclusions that function to order to serve personal inter- strations of self-serving bias, and concluded that the effect depends
preserve or increase self-esteem ests. Instead of being neutral on a desire to maintain self-esteem. The extent to which the cur-
observers of social events, we rent context makes self-esteem concerns salient should therefore
may sometimes interpret them in ways that suit us (or ways that determine the strength of the reported effect. However, compet-
suit our ingroup more generally; Islam & Hewstone, 1993, and see ing motivations such as self-presentation can also reduce self-
Chapter 14, this volume), allowing us to feel better about what has serving attributions. For example, we may be less inclined to take
happened. credit for positive outcomes in public settings, either because we
Let’s assume that you have just done well in an exam. Your don’t want to be seen to show off, or in order to avoid any em-
tendency may be to conclude that this reflects your innate ability barrassment at failing to live up to the unduly favourable image
(self-enhancing bias). However, if you do badly, perhaps you will that this would imply (e.g., Weary et al., 1982).
decide that the questions were unfair or unusually difficult, or that Abramson and colleagues’ (1978) attributional reformulation
the person sitting next to you in the examination hall was dis- of learned helplessness theory (see above) implies that depressed
tracting you by sharpening his pencils so noisily (self-protective bias). people adopt an attributional style that is the precise opposite of
More generally, you may be inclined to arrive at conclusions that the self-serving pattern (they take rather than disown responsibil-
maintain your positive self-image. ity for failure). Indeed, research suggests that simply being in a
One of the earliest demonstrations of self-serving bias was bad mood can reverse self-serving biases (e.g., Forgas, Bower &
an experiment by Johnson, Feigenbaum and Weiby (1964). Moylan, 1990), perhaps by removing the illusory glow that ordin-
Participants were educational psychology students and their task arily preserves our sense of well-being in happier states (Taylor &
was to teach two children how to multiply numbers via a one-way Brown, 1988, and see above).
intercom system, which meant that they never actually saw or
heard the children. The first phase involved explaining how to Motivational or cognitive effect? In the 1970s, a debate
multiply by 10 and the second phase involved explaining how arose about whether self-serving biases were genuinely self-
to multiply by 20. After each phase, the pupils’ worksheets were re- serving. Miller and Ross (1975) proposed that some personally
turned to participants, allowing them to assess how successfully advantageous attributions were entirely rational, while others
the concepts had been conveyed. In fact, the worksheets were con- simply reflected the application of principles of explanation that
cocted by the experimenters to manipulate whether or not the would normally be valid. According to this view, people do not dis-
answers were correct. In both conditions, pupil A answered the tort their thinking to protect self-esteem (motivational explanation)
questions on both worksheets correctly. However, pupil B either but rather use rules of thumb that happen to lead to faulty conclu-
did badly on both tasks or did badly on the first but improved on sions on some occasions (cognitive explanation). Take the educa-
the second. In other words, the students either failed or succeeded tional psychology students in Johnson and colleagues’ research. It
in teaching pupil B how to multiply. In the condition where pupil would be illogical for them to attribute pupil B’s improvement on
9781405124003_4_003.qxd 10/31/07 2:55 PM Page 62

62 CHAPTER 3 SOCIAL PERCEPTION AND ATTRIBUTION

the second exercise to the pupil’s abilities, because the pupil had done contributory factors. Instead, you look specifically for those causes
badly on the first sheet. Further, improvement followed the that will further your ends in the current situation. If the explana-
teacher’s careful attempt to explain 20 times multiplication tion is formulated while talking with a friend who is disappointed
after the pupil had done badly in 10 times multiplication. Under by her own performance, you might search your memory for any
these circumstances, it makes perfect sense to conclude that the bits of good luck that aided your success. If, on the other hand,
application of the participant’s teaching skills led to success. In the you are trying to score a point against the other person, you might
failure condition, by contrast, the pupil did not improve despite try to think of some particularly clever things that you wrote.
renewed efforts at explanation, and was consistently worse than Hilton (1990) argues that the explanations we provide in
pupil A on both exercise sheets. Thus, failure covaries with pupil conversations are specifically designed to meet the information
B but not with pupil A, or with the person doing the teaching. requirements of the person we are talking to. For example, when
According to Kelley’s covariation principle, even a detached discussing with my local greengrocer how I got sick after trying
observer should attribute bad performance to pupil B rather than kiwi fruit for the first time, I will tend to attribute the sickness to
the teacher when these conditions hold. the fruit, because I assume that she is interested in possible reac-
More generally, Miller and Ross argued that apparently self- tions to different fruits. However, when explaining my kiwi-induced
serving biases arise because effort covaries with success but not sickness to the doctor, her focus will be on what distinguishes me
with failure. If trying harder does not improve performance, then from other patients and what my particular complaint may be. In
it is reasonable to conclude that something about the task is pre- this context, therefore, I am more likely to attribute the sickness to
senting an obstacle. However, if trying harder does improve per- my own apparent allergy to kiwi fruit.
formance, then success is logically attributable to your trying. Note that these two explanations are mutually compatible and
Although these are valid points, few contemporary psycholo- may both be true, despite the fact that one refers to an external
gists would deny that thinking can also be distorted by motiva- cause (the kiwi fruit) and the other to an internal cause (my allergy).
tions and emotions. Indeed, the idea that we adjust our inferences Typically, a number of factors need to be in place to cause a given
to match existing positive expectations already sounds rather like event, and attribution involves selecting which of these factors to
an acknowledgement that we want to make ourselves look good emphasize in a particular context. To look at this another way, a
under certain circumstances (Tetlock & Levi, 1982). Thus, many number of changes in prior events could have averted the effect of
apparently cognitive explanations can be translated into motiva- getting sick (I could have eaten a different fruit, not had a kiwi fruit
tional terms, and many apparently motivational explanations can allergy, not liked the taste and refused to eat the fruit, and so on)
be translated into cognitive terms. Under these circumstances, and each of these possible changes reflects one of the causes
trying to tease apart cognitive and motivational processes is prac- contributing to my sickness (the fruit, my allergy, my liking of the
tically impossible. taste, etc.). Deciding which of these causes to emphasize depends
on what you think the person to whom you are explaining the
event already knows about its causes, and what you think they
The naïve scientist metaphor expected to happen.
On some occasions, of course, another person’s expectations
Do lay perceivers behave as scientists when making causal about what would normally happen are not entirely obvious, lead-
attributions, or do they have more practical concerns? ing to ambiguity about how to approach the explanatory task.
However, the precise phrasing of the causal question often helps
Most of the theory and research considered above assumes that to clarify matters. For example, if someone asks you ‘Does kiwi
people seek to understand the social world in a detached, scientific fruit make you sick?’ the question’s implicit emphasis (i.e., kiwi
manner, but sometimes get it wrong. This assumption is generally fruit as the subject of the action verb) may suggest that the ques-
known as the naïve scientist tioner wants to know about the fruit’s effects. According to Hilton
model (e.g., Fiske & Taylor, (1990), some apparent cases of bias can be explained by applying
naïve scientist model a metaphor for how
social information is processed that likens 1991), and, like all metaphors, these conversational principles. For example, Nisbett and col-
social perceivers to academic researchers it has its limits. Perhaps some leagues (1973) found that students explained their own choice of
who attempt to develop theories and of our explanations are not course in more situational terms than their best friend’s choice of
explanations for the purposes of prediction designed to provide a neutral course, consistent with the actor–observer difference. However,
and control of behaviour
characterization of reality in the emphasis of the question ‘why did you choose this university
the first place. In this case, course?’ naturally falls on the course as the topic about which in-
evaluating attributions against abstract rules of inference such formation is required. By contrast, the question ‘why did your best
as Kelley’s covariation principle is rather like complaining that friend choose this course?’ implies that the investigators want to
someone playing draughts is not correctly following the rules know about the friend rather than the course (otherwise, why not
of chess. directly ask for the participant’s own reasons?). In this study, then,
If people are not trying to be scientific when making attribu- the reported actor–observer difference may simply reflect a ration-
tions, what are they trying to do? One possibility is that explana- ally motivated attempt to provide the kind of information that
tions are generated to solve specific practical problems (White, was implicitly requested.
1989). For example, if you have to explain why you have done well An experiment by McGill (1989) supports this reasoning. She
in an examination, you probably don’t weigh up all possible found that a simple change in wording reversed the effect found
9781405124003_4_003.qxd 10/31/07 2:56 PM Page 63

ATTRIBUTION THEORY 63

by Nisbett et al. When participants were asked ‘why did you in

particular choose this course?’ they tended to offer less situational
explanations than when they were asked ‘why did your friend
choose this course in particular?’
More generally, attribution experiments may be viewed as con-
versations taking place between experimenters and participants in
which the participants are trying to work out what information
the experimenter is seeking (Schwarz, 1994). Making sense of ques-
tionnaire items often depends on thinking about who is asking
these questions and for what purposes. Adopting this kind of con-
versational approach also allows a reinterpretation of other sup-
posed illustrations of attributional bias. For example, in Jones and
Harris’s (1967) original demonstration of the correspondence bias
(discussed earlier), participants were told to work out another
student’s attitudes towards Castro. In order to do this, they were
provided with an excerpt from an essay that this other student Plate 3.6 Attributions do not always function as part of a
had (supposedly) written. Should participants simply ignore this cooperative process between people.
essay if its writer had been told what position to take on this issue?
Surely the experimenter wouldn’t have gone to the trouble of
showing it to them if it provided no information about the writer’s
opinions. Participants probably assume, therefore, that the experi- way that events are formulated already carries implications for
menter believes that the essay is relevant and conclude that they what caused what and who is to blame. Clearly, this isn’t a co-
are meant to infer the writer’s opinion on the basis of its content. operative process, but rather one in which attributions are actively
Indeed, when participants are explicitly warned that some of the contested.
material they will see may not be relevant, the correspondence Like White (1984), Edwards and Potter (1993) believe that
bias is reduced (Wright & Wells, 1988). attributions are formulated for particular purposes, but these pur-
poses specifically reflect conversational goals such as persuading,
undermining, blaming or accusing. Conversational maxims are
Attributions as discourse often explicitly flouted when explanations are presented. Further,
a range of alternative tactics and strategies may be improvised on-
Do attributions always function as part of a cooperative process line in response to the other person’s formulations. In this view,
between people? attributions do not function as attempts to explain a separately
existing social reality, but instead to construct a version of reality
Hilton’s working assumption is that people try to provide other suited to the current conversational business. Attributions are
people with information that helps to complete their understanding not descriptive representations but rhetorical moves in an ongoing
of events (Grice, 1975). However, conversations are not always dialogue.
cooperative processes in which information is generously exchanged.
Often, our aim is not to help someone else understand what has
happened, but rather to argue against them or defend our own
point of view against their attack. According to Edwards and
Potter (1993), attributions may be formulated rather differently
in these more antagonistic contexts. For example, they present a SUMMARY
transcript of a court case in which an allegation of rape is being
contested. The defence barrister is questioning the victim of the In this central part of the chapter we have covered a huge
alleged rape and apparently trying to suggest that she is partly amount of ground. We reviewed the major theories of
culpable. The interchange ran as follows: attribution and identified some of their limitations, includ-
ing the need to explain how general and specific knowledge
Barrister: (referring to a club where the defendant and the victim is used as part of the attribution process. We also consid-
met) It’s where girls and fellas meet isn’t it? ered applications of attribution theory, notably to clinical
Victim: People go there. (Edwards & Potter, 1993, p. 30) depression. Next we considered the key role of attributions
in emotion and the misattribution of arousal, and summar-
Note that the way the barrister describes what happened on the ized the evidence for the main types of attributional bias
night in question implies certain motives and intentions on the and their underlying causes. Finally, we evaluated whether
part of the victim, i.e., that she had gone to this place specifically lay perceivers do, in fact, function as naïve scientists, and
to meet with members of the opposite sex. Her response in turn is the role of causal attributions within the study of conversa-
designed specifically to neutralize this inference. Although neither tions and everyday discourse.
party to this exchange is explicitly presenting explanations, the
9781405124003_4_003.qxd 10/31/07 2:56 PM Page 64

64 CHAPTER 3 SOCIAL PERCEPTION AND ATTRIBUTION

how these two kinds of process – verbal representation as shaped

SOCIAL PERCEPTION AND by conversational pragmatics and rhetoric, and direct perception
determined by active pick-up of social information – might relate
SOCIAL REALITY to one another.

What are the main constraints on how social perceivers understand

other people and explain their actions? SUMMARY AND
CONCLUSIONS
Conversational and discursive models of attribution take us some
distance from Asch’s and Heider’s project of finding out how
people privately make sense of other people and what they do. l Person perception is influenced by the form as well as the
More generally, they unsettle the notion that people are simply content of information, and not all information is equally
naïve scientists trying to uncover the structure of an independent weighted.
social reality. Subsequent models have seen social perceivers l Some kinds of information (e.g., facial configuration) are
as lawyers (Hamilton, 1980), pragmatists (White, 1984), tacticians
perceived as direct indicators of personality, and some kinds
(Fiske & Taylor, 1991) or politicians (Tetlock, 2002). Although
of information are weighted highly (e.g., first-presented
there is some truth in all of these characterizations, the bottom
information and ‘central’ traits) but rarely in all situations.
line seems to be that a variety of strategies are available for deal-
ing with social information, which may be deployed selectively l Causal attribution is shaped by prior general knowledge, as in
depending on circumstances. correspondent inference theory and covariation theory. But
But does this mean that social perceivers are free to construct we do not always have access to this information and specific
whatever formulation suits their current purposes? There are three knowledge is also used.
reasons why this is not the case. The first is that biology and cul- l Inferences about the causes of achievement influence
ture do not equip people with infinitely flexible conceptual re-
motivation and, relatedly, internal, stable and global
sources for understanding other people. On the one hand, we are
attributions for helplessness may exacerbate depression.
innately attuned to certain kinds of social information at the
expense of others (e.g., Fantz, 1963; Johnson & Morton, 1991). l Attributions for internal symptoms may alter interpretations
On the other hand, we are socialized into particular ways of think- of emotional experience.
ing about the social world (e.g., social representations; Farr &
l As with person perception, not all information is equally
Moscovici, 1984). Attribution and social perception always take
weighted when making attributions about the causes of
place against the backdrop of norms of understanding that make
behaviour; causal information that is salient is especially
some inferences more likely than others.
influential.
A second and related point is that other people will contest any
formulation of social reality that doesn’t match their own ( just as l Various biases have been identified, which qualify the general
their own formulation may be contested in turn). The upshot is theories and bring them more into line with how attributions
that some consensus tends to emerge among people who have operate in everday life.
regular contact with one another. However, social reality may also
l People in individualistic societies tend to overestimate
be influenced by the representations that are applied to it.
personal causes of behaviour (the correspondence bias),
For example, our judgements about others can lead us to behave
but this correspondence bias is neither inevitable nor
in ways that bring out the very characteristics that we expected
uncontrollable.
(via self-fulfilling prophecies; see Snyder, 1984, and Chapter 10,
this volume). l Differences in actors’ and observers’ attributions depend
The final constraint on representations is the content of partly, but not entirely, on their different perceptual
the social information itself. Although social perception research perspectives.
tends to focus on the interpretation of verbally represented infor-
mation or static, sensory stimuli, when people confront each other
l Motivational factors are implicated in some instances of
in everyday life they often have access to a dynamic multimodal self-serving bias.
presentation that is responsive to their own conduct. Some char- l Attribution typically operates within a conversational
acteristics of others can be read directly from the available infor- context and is responsive to conversational demands.
mation (Baron & Boudreau, 1987; Gibson, 1979). For example, we
register where someone else’s attention is focused from the ori-
l Although people sometimes act like naïve scientists, they
entation of their sensory organs. It seems therefore that some kinds make attributions in ways that are strategic for the goals
of social perception and attribution are not explicit verbally medi- of social interaction, serving a variety of rhetorical
ated processes but instead involve direct registration of sensory purposes.
information. The challenge facing future research is to specify
9781405124003_4_003.qxd 10/31/07 2:56 PM Page 65

SUGGESTIONS FOR FURTHER READING 65

Suggestions for further reading the field, including intrapersonal, interpersonal, intergroup
and societal aspects of attribution.
Ross, L. & Nisbett, R.E. (1991). The person and the situation:
Fiske, S.T. & Taylor, S.E. (1991). Social cognition (2nd edn). New
Perspectives of social psychology. Boston: McGraw-Hill. Highly
York: McGraw-Hill. Includes treatment of attribution theory
readable introduction to the cognitive perspective on
(Chapters 2 and 3) within an overarching perspective on
attributional bias and other aspects of social perception and
social cognition.
inference.
Försterling, F. (2001). Attribution: An introduction to theories,
Zebrowitz, L.A. (1990). Social perception. Buckingham: Open
research, and applications. Hove: Psychology Press. An
University Press. A thorough review of research into social
accessible overview of attribution models and findings from
perception that attempts to integrate ecological and cognitive
Heider to the early twenty-first century.
approaches.
Hewstone, M. (1989). Causal attribution: From cognitive processes to
collective beliefs. Oxford: Blackwell. A wide-ranging view of