what makes people angry? laypersons' and psychologists' categorisations of anger in the...
TRANSCRIPT
This article was downloaded by: [UQ Library]On: 15 November 2014, At: 18:45Publisher: RoutledgeInforma Ltd Registered in England and Wales RegisteredNumber: 1072954 Registered office: Mortimer House, 37-41Mortimer Street, London W1T 3JH, UK
Cognition andEmotionPublication details, includinginstructions for authors andsubscription information:http://www.tandfonline.com/loi/pcem20
What Makes PeopleAngry? Laypersons'and Psychologists'Categorisations ofAnger in the FamilySandra Carpenter & Amy G.HalberstadtPublished online: 10 Sep 2010.
To cite this article: Sandra Carpenter & Amy G. Halberstadt(1996) What Makes People Angry? Laypersons' and Psychologists'Categorisations of Anger in the Family, Cognition and Emotion, 10:6,627-656, DOI: 10.1080/026999396380088
To link to this article: http://dx.doi.org/10.1080/026999396380088
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracyof all the information (the “Content”) contained in thepublications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or
warranties whatsoever as to the accuracy, completeness,or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions andviews of the authors, and are not the views of or endorsedby Taylor & Francis. The accuracy of the Content should notbe relied upon and should be independently verified withprimary sources of information. Taylor and Francis shall not beliable for any losses, actions, claims, proceedings, demands,costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connectionwith, in relation to or arising out of the use of the Content.
This article may be used for research, teaching, and privatestudy purposes. Any substantial or systematic reproduction,redistribution, reselling, loan, sub-licensing, systematic supply,or distribution in any form to anyone is expressly forbidden.Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
What Makes People Angry?
Laypersons’ and Psychologists’
Categorisations of Anger in the Family
Sandra Carpenter
The University of Alabama in Huntsville, USA
Amy G. Halberstadt
North Carolina State University, USA
People’ s categorisations of anger causes were examined in these studies, in
order to obtain a layperson taxonomy of anger causes in families, and to
compare this taxonomy to those created by psychologists. In Study 1, college
students sorted descriptions of anger-eliciting events in family relationships.Hierarchical cluster analyses of their classi ® cations resulted in a total of 15
separate categories of causes for the three types of familial relationships.
These categories conformed to a prototypical format and varied in interesting
ways across relationship and gender. Study 2 compared the students’ classi-
® cation schemes to ® ve different schemes developed by psychologists.
Considerable overlap in students’ and psychologists’ categorisations was
demonstrated for some categories of causes, but not for others. Study 3
examined the three dimensions along which layperson and psychologist
classi ® cations might differ. The dimensions of speci ® c versus abstract andexperiential versus scienti ® c seemed to account for some of the differences
between psychologists’ and laypersons’ classi ® cations found in Study 2. The
dimension of description versus explanation did not account for differences,
but may account for similarities in the two types of classi ® cations. We
suggest that people’ s perceptions of causes better capture the phenomenal
experience of anger, whereas psychologists’ theoretical models of anger
better capture the underlying structures.
COGNITION AND EMOTION, 1996, 10 (6), 627± 656
Requests for reprints should be sent to Sandra Carpenter, Department of Psychology, The
University of Alabama, Huntsville , AL 35899, USA. Email: [email protected]
Support for this research was provided to Sandra Carpenter by a grant from The University
of Alabama in Huntsville, and to Amy Halberstadt by a NIMH grant (No. 1 R03 MH42425).
Portions of this manuscript were presented as a paper, Anger within family relationships, at
the 1993 meeting of the Society for Research on Child Development, held in New Orleans,
Louisiana. Other portions were presented as a paper, Categories of anger in relationships, at
the 1992 meeting of the Society for Experimental Social Psychology, held in San Antonio,
Texas. We thank our coders: Susan Trentham, Eileen Sampson, Jean Haake, Terry Sterry,
and Jennifer Lacher-Smith for help in classifying the data.
q 1996 Psychology Press, an imprint of Erlbaum (UK) Taylor & Francis Ltd
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
628 CAR PENTER AN D HALBERSTA DT
INTRODUCTION
Anger is an ubiquitous emotion. Anger was the emotion most frequently
reported when people were asked to describe their most recent emotional
experience (Scherer & Tannenbaum, 1986) . Research focused speci® cally
on social anger (that is, anger directed at other persons) found that 80 ± 83%
of respondents reported feeling angry at least once a week (Averill, 1982;
Halberstadt & Carpenter, 1994). Because of the frequency with which
anger is experienced and the important interpersonal consequences that
follow from its expression, an understanding of the causes of anger has
both theoretical and practical usefulness. Further, `̀ good science’ ’ dictates
that clear and extensive description of a phenomenon is prerequisite to
effective theory-build ing. The present research develops a layperson’ s
taxonomy of the causes of anger, in order to facilitate such theory-build-
ing. The empirically obtained taxonomy is then analysed in terms of three
theoretical issues. (1) Are causes of anger socially constructed, such that
they differ across types of interpersonal relationships ? (2) Are causes of
anger cognitively represented as prototypes? (3) Does the layperson tax-
onomy differ from psychologis ts’ previously developed taxonomies of
anger causes, and in what ways? Answers to these questions will provide
us with a strong basis for developing new theories of why people become
angry and in what relationships .
Several taxonomies of anger have been generated in the last 10 years or
so, usually within the context of identifying a prototype for anger or
creating a universal taxonomy for all emotional experience (Averill ,
1982; Camras & Allison, 1989; Fisher & Johnson, 1990; Fisher, Reid, &
Melendez, 1989; Scherer & Tannenbaum, 1986; Scherer, Wallbott, &
Summer® eld, 1986; Shaver, Schwartz, Kirson, & O’ Connor, 1987) . How-
ever, if anger is considered to be socially constructed, causes of anger
should differ as a function of relations in `̀ social structures’ ’ (e.g. power
and interdependence). The degree to which events elicit anger may there-
fore depend on recipient and instigator characteristics, such as gender and
age. Thus, because the experience of anger may be affected by the social
context, speci® cally by the relationship between the interactants, it is
important to create taxonomies that re¯ ect the unique structures of differ-
ent relationships.
The family provides an excellent opportunity to examine perceived
causes of anger as a function of relationship differences. Researchers
who have begun to study anger in familial relationships propose that
many episodes of anger are due to inabiliti es to conform to the norms or
rules that family members expect each other to accept and to honour
(Fisher & Johnson, 1990; Fisher et al., 1989), and it makes sense that
familial expectations differ as a function of the type of relationship (i.e.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
parent to parent, parent to child, child to parent). For example, children are
considered to be less `̀ in control ’ ’ of their behaviours than are adults.
Perhaps because we tend to be less angry at those who cannot control
their behaviours, anger is often less extreme toward children than toward
adults (Savitzky, Czyzewski, Dubord, & Kaminsky, 1975). Family rela-
tionships also vary as a function of age and power, and age and power
differences may create different types of anger experiences (Averill , 1982).
Therefore, we predicted that different causes of anger will predominate for
different types of familial relationships .
In addition, we propose to study causes of anger in terms of their
cognitive representations . Several theorists and researchers (Fehr, Rus-
sell, & Ward, 1982; Russell, 1991; Shaver et al., 1987) have suggested
that individual emotions are cognitively represented as prototypes. Russell
(1991) proposes that emotions can be conceived of as prototypes because
instances of emotions: (a) are organised into differing categories (e.g. fear,
sadness, happiness, anger), (b) are hierarchically structured (e.g. fury and
rage are subordinate types of anger, which in turn is a subordinate type of
emotion), and (c) have overlapping , fuzzy boundaries (fear and anger could
coincide, for example). However, emotions also share several compo-
nents Ð antecedents, physical signs, behavioural consequences, and influ-
ence on perceptions and thoughts Ð and these may themselves be organised
in terms of a prototype. Thus, we predicted that one component of the
prototype of angerÐ the antecedent or `̀ cause’ ’ Ð is, itself, represented as a
prototype. To test this, the three criteria just mentioned will be applied to
classi® cations of the causes of anger.
Theorists who have attempted to delineate the prototype of anger have
typically developed categorisation schemes by perusing the responses of
their participants, creating categories on the basis of these responses, then
having research assistants code the responses into categories (Averill ,
1982; Camras & Allison, 1989; Scherer & Tannenbaum, 1986; Shaver et
al., 1987). Averill (1983) and Anderson (1991) suggest, however, that the
`̀ person in the street’ ’ may have different conceptions of the causes of their
anger than do psychologis ts, even though psychologica l theories are based
on the self-reports of those persons. Averill and Anderson provide evidence
that laypersons’ knowledge is phenomenal, intuitive, and unsystematic,
whereas psychologists, as scientists, work toward making that knowledge
explicit in scienti® c descriptions . Anderson speci® es that laypersons’ and
theorists’ descriptions may differ in terms of the type of description given
(phenomenal experience vs. underlying structure), how the description is
organised (categorical vs. dimensional) , and how generalisable the descrip-
tion is (what people typically do vs. what people can do). The degree to
which people’ s representations of the causes of anger match psychologists ’
theories of anger causes has yet to be ascertained. Thus, our third purpose
W HAT M AKES PEO PLE AN G RY ? 629D
ownl
oade
d by
[U
Q L
ibra
ry]
at 1
8:45
15
Nov
embe
r 20
14
630 CAR PENTER AN D HALBERSTA DT
was to compare our obtained layperson taxonomy to categorisations
derived by psychologists. We expected that laypersons’ and psycholo-
gists’ taxonomies of anger would differ, as suggested by Anderson and
by Averill.
Study 1 was designed to determine how laypersons categorise causes of
anger within different relationships, and whether these anger causes are
cognitively represented as prototypes. Study 2 was designed to determine
the similarities and differences between the obtained layperson categorisa-
tions and psychologists’ taxonomies created for this purpose. When differ-
ences emerged in Study 2, Study 3 was designed to identify dimensions
that might underly differences between psychologists’ and laypersons’
taxonomies.
STUDY 1
Method
The method of Study 1 required participants (mothers) to generate descrip-
tions of causes (of anger) across a range of situations (relationship types).
These causal descriptions were then sorted into categories by an indepen-
dent group of participants (students) who were given no a priori structure
for categorisation. Rather, they were allowed to categorise the stimuli in
any ways they wished. This sorting technique has previously been used in
several studies to investigate cognitive representations. For example,
researchers have used this technique to ascertain the perceived causes of
emotions (Scherer, 1984; Shaver et al., 1987), the representational structure
of causes (Anderson, 1991), the perceived importance of gender and age as
categories in person perception (Brewer & Lui, 1989) , and the subjective
organisation s of lexical information (Miller, 1969).
Participants
Descriptions of anger causes were generated by 60 mothers from the
environs of Washington, DC, who had at least one child between the
ages of 5 and 7 years participating in another study (Halberstadt, Fox, &
Jones, 1993). The majority were married (48 married, 5 remarried, 6
divorced or separated, 1 never married), and they had one to nine children
(a median of two). Their mean age was 35.5 years. The sample was 78%
white, 17% African-American, and 5% other racial/ethnic groups.
Mothers’ descriptions of anger causes were sorted by 59 female and 68
male college students who participated for course credit. The vast majority
of the sample had never been married (56 of the males and 36 of the
females) and had no children (61 of the males and 49 of the females). Their
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 631
mean age was 25.37 years. Students at this southeastern university are
primarily white (85%), with minorities of African-American (9%) and
other racial/cultural groups (6%).
Materials
The responses of the mothers were used as the stimuli in this study. The
reliability and validity of mothers’ reports of their own anger and anger of
family members have been documented elsewhere (Carpenter & Halber-
stadt, submitted).1
The six relationships explored, and the number of different items
generated for each relationship, were as follows: (a) mother’ s anger at
father: 54 items, (b) father’ s anger at mother: 52 items, (c) mother’ s anger
at child: 45 items, (d) father’ s anger at child: 41 items, (e) child’ s anger at
mother: 47 items, and (f ) child’ s anger at father: 44 items.2
Each of the
items was typed on to a separate strip of 8.5inch 3 2inch paper, which was
then laminated.
1Stringent rules for establishing temporal reliability and consensual validity by parents
about anger causes were used. Items were coded dichotomously : as agreements if the
responses were synonyms (i.e. `̀ money troubles’’ and `̀ ® nancial matters’’ ); but as disagree-
ments if one response differed in content or level of speci® city (e.g. mother mad at father for
his insults toward her vs. his `̀ disrespect’ ’ toward her). As a reliability check, a graduate
student blind to the hypotheses of this study coded all items for two of the eight relation-
ships; agreement with the ® rst coder was 100%. The degree to which mothers provided
exactly the same anger event description across the two-week period ranged from 51% (for
child’ s anger at father) to 78% (for sibling anger at target child) across the eight relationships
(M = 60%). These percentages need to be interpreted within the context of an almost in® nite
number of possible speci ® c causes. For example, in the sample, 52 mothers generated 35
different reasons for anger at their spouses in the ® rst session and an additional 15 in the
second session. Thus, given these probabiliti es, mothers’ perceptions of what makes family
members angry at one another showed a good deal of consistency over time.
To ascertain accuracy of mothers’ perceptions, 31 mothers’ descriptions were compared to
those of fathers in the same family. The degree to which fathers provided exactly the same
anger description as mothers ranged from 42% (for sibling anger at target child) to 67% (for
child anger at father) across the eight relationships (M = 56%). Again, these percentages
need to be interpreted within a context of parents generating 33 possible reasons for mothers
to be angry at fathers.2
Some of the mothers generated identical responses (e.g. seven mothers listed `̀ doesn’ t
listen’ ’ for the `̀ mother’ s anger at child’ ’ set). In these cases, only one such stimulus item was
presented to the Study 1 participants. However, even minimally varying responses of the
mothers were considered different and were individually presented to participants. Also,
some divorced mothers omitted certain items relating to father± child anger when they felt
that they were no longer familiar with those relationships .
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
632 CAR PENTER AN D HALBERSTA DT
Procedure
Participants were run in groups of four to six persons. They were given the
six sets of mothers’ descriptions of reasons for anger in the family and were
instructed to sort each set of items into more than two piles, `̀ so that
similar causes for anger end up in the same pile together and different
causes for anger end up in different piles’ ’ . The six sets of anger descrip-
tions were presented to participants in individuali sed random orders.
Additionall y, the items within each set were randomly arranged. Whereas
some participants were able to complete all six sorts in the one-hour
session, other participants could complete only four or ® ve sets. There-
fore, different numbers of participants completed the six sorts: (1)
`̀ mother’ s anger at father’ ’ was sorted by 109 participants; (2) `̀ father’ s
anger at mother’ ’ was sorted by 112 participants; (3) `̀ mother’ s anger at
child’ ’ was sorted by 100 participants; (4) `̀ father’ s anger at child’ ’ was
sorted by 106 participants; (5) `̀ child’ s anger at father’ ’ was sorted by 107
participants, and (6) `̀ childs ’ anger at mother’ ’ was sorted by 112 partici-
pants. Thus, the range of the number of participants completing the sorts
was 100 ± 112.
Results
The number of categories that participants used for sorting the items ranged
from 3 to 17. For each of the six sets of anger causes, the incidence with
which participants sorted items into the same pile was recorded in a co-
occurrence matrix. For each of the six sets, a square matrix (ranging from
41 3 41 items to 54 3 54 items) indicated the number of partic ipants who
sorted each pair of the items into a set together. Each of the six matrices
was submitted to a cluster analysi s using the average linkage method
(similar to Shaver et al., 1987). Each cluster analysi s indicated the items
of a set that were typically or often grouped together by participants. Items
were considered members of a particular group if the distances to items
within the group were smaller than distances to items in other groups. From
each cluster analysis, the categories that participants used to sort the
stimuli could be inferred. These are reported as follows.
Parents’ Anger at Parents. As shown in Table 1, the cluster analyses
of mothers’ reports about mother’ s anger at father and father’ s anger at
mother resulted in six and ® ve clusters, respectively. There were substan-
tial similarities in the clusters that emerged; both mothers and fathers were
reported as angry about: (1) Money Issues; (2) Children Issues; (3) Person-
ality De® ciencies; and (4) Inadequate Time, although cluster size did vary
by parent gender.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 633
It is notable that the women who generated these items considered their
own personality de ® cienc ies to be the predominant cause of fathers’ anger,
but not vice versa. It may be that mothers are perceiving in their spouses
the fundamental attribution bias, but somehow avoiding it themselves; they
report that fathers see mothers as having personality characteristics that
cause fathers’ anger, rather than seeing the anger as contextually or
dyadically based (which is more how mothers appear to be reporting the
causes of their own anger). These results suggest that mothers may have a
more transactional or systems approach to the situation. That is, mothers
may recognise the importance of the interaction between the two interact-
ing partners, whereas they perceive fathers as placing causes for events as
TABLE 1
Layp erson Anger C ategories for Paren ts’ Anger at Each O ther and Exam ples o f Item s
C lu stered in to those Categ ories
Parental Anger at Each Other
What m akes m others angry at fathers What m akes fathers angry at m other s
Money Issues (n = 5) Money Issues (n = 7)
Spending money and time Financial matters
His job Shopping too much
Spending money like water Inability to keep track of the cheque book
Children Issues (n = 7) Children Issues (n = 4)
Lack of participation with kids Wants to eliminate milk from daughter’ s diet
Spoils daughter Disagrees about handling of children
Impatience with children Allowing children to misplace things
Personality De® ciencies (n = 6) Personality De® ciencies (n = 19)
Short temper Impatience
Jealousy Compulsivity
Sulky behaviour Too honest
Inadequate Time (n = 6) Inadequate Time (n = 13)
Watching TV She works late
Business trips he might not have to take Doesn’ t spend enough time alone with him
Spending time with friends, not family Lack of sex
Not Being a Caring Partner (n = 16) Being Controlling (n = 9)
Coming home late without calling Her nagging him
Lack of communication She has too many demands
Lack of interest/involvement Assigning him excessive duties
Not Contributing to Households (n = 14)
Neglect of duty
Asking to ® nd something
Slow to help
Note : n indicates the number of items sorted into that category. Examples are abbreviated
descriptions of mother-generated causes.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
634 CAR PENTER AN D HALBERSTA DT
more `̀ in one person’ ’ . Although conjectural at this point, these possible
differences are surprisingly related to other gender differences in the
communication and socialisation literatures (e.g. Aries, 1987; Maltz &
Borker, 1982), and are worth pursuing in future research.
Another difference in the sorts is how the issue of inadequate time and
attention is treated. Whereas fathers’ anger about time and attention is
represented in only one category (Inadequate Time: `̀ she works late’ ’ ,
`̀ lack of sex’ ’ ), mothers’ anger was classi ® ed into other-focused activities
(i.e. Inadequate Time: `̀ watching TV’ ’ , `̀ spending time with friends, not
family’ ’ ) and lack of interpersonal interest or communication (i.e. Not
Being A Caring Partner: `̀ coming home late without calling ’ ’ , `̀ lack of
interest’ ’ ). Thus, mothers may perceive themselves as more likely to
become angry due to a partner’ s lack of family involvement than are
fathers (for mothers, n = 22; for fathers, n = 13).3
Two other categories were unique to one parent. Only mothers’ anger
was sorted for Not Contributing To Household (n = 14). A perusal of
fathers’ items identi ® ed only two similar items (`̀ inability to keep track
of the cheque book’ ’ and `̀ not getting things done’ ’ ), and these were sorted
into the Money Issues and Personality De® ciencies categories. Also, only
fathers’ anger was sorted for Being Controlling (n = 9). A perusal of
mothers’ items identi ® ed only one similar item (`too directive and dicta-
torial ’ ’ ), and this was sorted into the Personality De® ciencies category. The
Not Contributing to Household and Being Controlling categories were
comprised mostly of items expressing, respectively, mothers’ concerns
that they are carrying too much of the burden of household chores, and
fathers’ concerns that their wives are playing the `̀ heavy’ ’ and demanding
too much of them, mostly with regard to household work. Thus, in many
households, the negotiation, or lack of it, regarding completion of house-
hold tasks appears to be a prevalent source of tension.
Parents’ Anger at Children. As shown in Table 2, the cluster analyses
of the sorts for maternal report about mothers’ anger at children and
fathers’ anger at children each resulted in four clusters. There were sub-
stantial similarities for three of these clusters; both mothers’ and fathers’
anger causes were classi ® ed as: (1) Disobedience/Not Listening; (2)
3Although chi-square analyses might be desirable for comparing the frequency of a causal
category across relationships , the categories that emerged from the cluster analyses were not
identical across all relationships . Consequently, item frequencies across categories could not
be compared fairly (the frequency could only be zero when items were not sorted into that
category for that relationship) . Now that these categories have been generated, future
research might include classi® cation across all categories so that differences between causes
of spouses’ anger, for example, can be statistically analysed.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 635
Naughty Behaviours ; and (3) Personality De® ciencies. However, one
interesting difference also emerged; the categorisations of damage beha-
viours that made mothers angry were predominantly interpersonal, whereas
categorisations of fathers’ reported anger focused on damage to property.
Children’ s Anger at Parents. As shown in Table 3, the cluster analyses
of maternal reports about children’ s anger at mothers and children’ s anger
at fathers yielded ® ve and four clusters, respectively. Again, there were
substantial similarities for four clusters; children were reported to be angry
at both mothers and fathers for: (1) Inadequate Attention; (2) Discipline;
(3) Unfair Treatment; and (4) Not Getting One’ s Way. Only mothers’ anger
was sorted for Power Struggles, and a perusal of the set of fathers’ items
identi ® ed only three similar items (`̀ power struggles’ ’ , `̀ makes child do for
self ’ ’ , and `̀ disagreements about plans’ ’ ). These three items were sorted
into the Unfair Treatment category. Although discipline seems to charac-
terise more of children’ s reported anger at father (n = 10) than at mother (n
= 4), mothers’ additional category of Power Struggles (n = 11) suggests
that discipline is not any less of a cause for anger toward mothers, but is
perceived differently.
TABLE 2
Layp erson Anger C ategories for Parents’ Anger at Ch ild ren an d Exam ples o f Item s
C lu stered in to those Categ ories
Parental Anger at Children
What m akes m others angry at ch ild ren What m akes fathe rs angry at ch ild ren
Disobedience/Not Responding (n = 14) Disobedience/Not Listening (n = 16)
Not listening Acting smart
Not following directions Won’ t follow directions
Having to ask her to do things again Refuses to do what she’ s asked
Naughty Behaviours (n = 13) Naughty Behaviours (n = 12)
Being rude Bad table manners
Acting like a baby Sucking her ® nger
Talking back Toys and clothes on the ¯ oor
Personality De® ciencies (n = 11) Personality De® c iencies (n = 7)
Dawdling Laziness
Slow to complete tasks Being slow
Bedwetting Being careless
Interpersonal Harm (n = 7) Damaging Behaviours (n = 6)
Fighting with brother Breaking something
Arguing with sibling Jumping on bed
Hurting her sister Playing too rough
Note : n indicates the number of items sorted into that category. Examples are abbreviated
descriptions of mother-generated causes.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
636 CAR PENTER AN D HALBERSTA DT
We speculate that these differences in anger toward parents may be due
to different parenting styles of mothers and fathers. Whereas fathers may
punish children for transgressions, mothers may be more likely to recog-
nise children’ s rights to disagree (resulting in power struggles). As can be
seen in Table 3, the causes listed by mothers about their own receipt of
anger include more negotiation by the child (e.g. decision child disagrees
with, child does not want to obey, child wants to assert her independence or
manipulate, child made to do for self ) as compared to father’ s punishment
(e.g. getting sent to room, told to obey, being too harsh). Again, there is a
hint of mothers’ greater transactional focus, with discipline seen as a more
dyadic process rather than an unilateral event. Mothers may be shifting
some of the power and rights to the child earlier than fathers, who are
described more as maintaining authority and power without sharing it with
children.
TABLE 3
Layperson Anger C ategories for Ch ild ren ’s Anger at Parents and Exam ples of Item s
C lu stered in to those Categ ories
Children’ s Anger at Parents
What m akes child ren angry at m others What m akes ch ild ren angry at fathers
Inadequate Attention (n = 8) Inadequate Attention (n = 8)
Not spending special time alone Dad travels too much
Failure to keep promises Not listening
Not getting immediate attention Won’ t play now
Discipline (n = 4) Discipline (n = 10)
Spanking Yelling
Correction Sends to room
Discipline Punishment
Unfair Treatment (n = 10) Unfair Treatment (n = 12)
Mom yelling Won’ t follow through with promises
Mom being impatient, rushing Disappointment
Mom interrupting child Child asked to hurry
Not Getting One’ s Way (n = 14) Not Getting One’ s Way (n = 14)
Refusal to buy something Not giving candy or toy
Not caving in Dad says `̀ no’ ’
Won’ t let him have his own way Not letting her have her way
Power Struggles (n = 11)
Decision child disagrees with
Child wants to assert her independence
Child doesn’ t want to obey
Note : n indicates the number of items sorted into that category. Examples are abbreviated
descriptions of mother-generated causes.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 637
Similarities and Differences across Relationships. This ® rst study has
provided us with important descriptive information about various types of
anger causes across relationships. The cluster analyses of participants ’
sortings of the stimuli resulted in 15 separate categories, with only two
categories reoccurring across relationships . Thus, as predicted, categories
were differentially relevant to the particular role relationship under con-
sideration (i.e. who is being angered by whom). Given that the reports of
anger are all from the perspective of one informant (the mother), such that
causes might be expected to be spuriously similar over relationships , these
differences are particularly striking.
The categories that do recur are also of interest. Inadequate Attention
was found for both the parents to parents and children to parents relation-
ships. The parents to children relationship also had a category of `̀ not
paying attention’ ’ (Disobedience/Not Listening), but this category was
more about the speci® cs of compliance, whereas the basic theme in the
other two relationships was that the parent or spouse didn’ t have enough
time for the other person. The second re-occurring category, Personality
De® ciencies, emerged for both the parents to parents (n = 25) and the
parents to children (n = 18) relationships, but not for the children to parents
relationship. Either children at this age do not succumb to the fundamental
attribution error, or are not perceived by others to be making stable
attributions of a dispositional nature.
Discussion
A taxonomy of perceived anger causes was developed in Study 1, and this
taxonomy revealed both the importance of social context and the proto-
typicality of anger causes. These descriptive data are important for our
understanding of how people think about anger causes. Whereas the
cognitive structure, in all cases, was that of a prototype, the content of
anger causes was found to vary across relationships .
First, as predicted, anger causes differed by relationship, with only two
overlapping categories. We have therefore identi ® ed divergent patterns of
anger causes that should be useful for both basic and applied (e.g. therapy)
psychological understanding . Some gender differences also emerged, with
slight category differences in each relationship type; these differences are
interpretable within the societal context of females’ and males’ roles as
spouses and parents. It would be interesting to explore whether the more
frequently used categories of anger causes apply only to relationships that
are intimate in nature or also apply to other relationships , such as those
with co-workers, neighbours , congressional representatives, etc. A next
logical step in categorising laypersons’ representations of the causes of
anger would be to add to the types of relationships investigated. Our
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
638 CAR PENTER AN D HALBERSTA DT
results, then, indicate that attempts at generic analysis of anger causes may
miss fundamental differences in the social construction of anger across
relationships .
Second, the cluster analysi s of anger descriptions yielded data that ® t a
prototype model, in that (a) the causes of anger were organised into varying
classi® cations, (b) the categories occurred at varying levels of speci® city
that differed across relationships (i.e. money issues vs. personality defi-
ciencies), and (c) the categories included particular instances of anger that
were not unanimousl y sorted by participants into any given category.
These, respectively, indicate (a) differing categories, (b) a hierarchical
structure, and (c) fuzzy boundari es between categories. Thus, people’ s
representations of anger causes ful ® l the three criteria used to determine
prototype representation. From a cognitive perspective, these results indi-
cate that causes of anger, as conceptualised by laypersons, although
different in content across relationship, are represented in similar ways.
The third issue we wished to address in our studies, with respect to
laypersons’ representations of anger causes, was: Do these categorisations
correspond to categories proposed by anger researchers? Because psychol-
ogists ’ coding schemes have been based on laypersons’ accounts of anger
elicitors, we expected at least some degree of similarity, even though some
coding schemes for anger were not created speci® cally for anger within
families (Scherer & Tannenbaum, 1986; Scherer et al., 1986; Shaver et al.,
1987) . However, it was predicted that laypersons’ and psychologists ’
taxonomies would differ as suggested by Averill (1983) and Anderson
(1991).
Knowledge of how these taxonomies differ is important in both basic
research and applied settings. In terms of basic research, it is useful to
identify similarities and differences in how `̀ novices’ ’ and `̀ experts’ ’ differ
in their classi ® cations. In terms of application, a taxonomy might better
enable a therapist to translate layperson categories to psychologists ’
theories and vice versa. Thus, the degree of overlap between laypersons’
and psychologists ’ categorisations of the causes of anger is examined in
Study 2, and the dimensions along which laypersons and psychologists
differ are examined in Study 3.
STUDY 2
Several researchers interested in the causes of anger have developed coding
schemes for categorising the types of events that elicit anger. Five such
coding schemes are compared to the laypersons’ categorisations generated
in Study 1. These categorisation schemes are described in their order of
publication.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 639
First, Scherer et al. (1986) telephoned 779 participants in eight countries
and asked them to describe their most recent emotional experiences for
four different emotions. From their codes, we chose only those relevant to
interpersonal relationships, and combined some categories (e.g. temporary
and permanent separation) . The six resulting codes described the causes of
anger as due to: (1) failure of others to conform to social norms, or failure
to be considerate about persons and property; (2) failure to reach goals or to
achieve an objective; or an unexpected, unnecessary inconvenience, or
time loss; (3) inappropriate rewards to self (i.e. being unjustly treated);
(4) separation, either temporary or permanent; (5) damage to personal
property by others and by oneself; and (6) physical hurt.
Second, Shaver et al. (1987) asked 120 college students to write about
speci® c or typical instances of emotional experiences. From these descrip-
tions, four categories of causes that discriminate types of anger were
developed: (1) violation of an expectation; things not working out as
planned; (2) frustration or interruption of a goal-directed activity; (3)
reversal or sudden loss of power, status, or respect; or insult; and (4)
physical or psychologica l pain, either real or threatened.
Third, Camras and Allison (1989) asked 47 adult women and 43 children
to write or tell stories about situations in which six kinds of emotions
would be felt. Twenty-four categories of causes were generated overall;
® ve categories were used for more than one anger story. These ® ve
categories were: (1) violation of a personal or social contract or an implicit
social rule; (2) not receiving a desired or desirable object, response, or
experience; (3) damage or attempts to damage a person’ s body, posses-
sions, self-esteem, or free agency (the action is not justi ® ed by the person’ s
behaviour or the authority of the agent over the person); (4) damage or
attempts to damage a person’ s body, possessions, self-esteem, or free
agency (the action is justi ® ed by the person’ s behavior or the authority
of the agent over the person); and (5) taking something away from a person
without justi ® cation.
Fourth, Fisher and Johnson (1990) asked 192 2nd-, 5th-, and 8th-grade
children, and college students to tell stories about situations that cause
parents or children to be angry at each other. Fisher and Johnson, on the
basis of the theoretical model of Bengston and Black (1973), predicted that
the causes of anger within a family would be related to con¯ icts arising
from these negotiations as well as to general ability in social perspective-
taking. Their six categories of anger causes were: (1) not ful ® lling role
expectations, de ® ned as a pattern of failure to conform to expectations now
and probably also in the future; (2) not ful ® lling social obligations , (i.e.
behaviours that threaten family functioning as a social unit; (3) not
helping (i.e. failure to provide instrumental or material assistance); (4)
failure to provide affection; nonresponsiveness to feelings; lack of
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
640 CAR PENTER AN D HALBERSTA DT
others’ empathy, liking , fairness, respect, love, or trust; (5) lack of con-
sensus in decision-making or values; and (6) low association or infre-
quency of contact.
Finally , Carpenter and Halberstadt (submitted) developed a taxonomy
based on the preceding four coding schemes, as applied to the descriptions
of anger in families that were generated by the 60 mothers reported earlier.
The best qualities of the four coding schemes (as determined through
theoretical considerations and empirical evidence) resulted in a 2 3 2 3
2 factorial design. (1) Type of Cause: expectancy violations vs. goal
blockage. (2) Focus of anger: person-focused vs. object-associated causes
of anger. (3) Temporal Speci ® city of cause: general, ongoing actions vs.
speci® c, circumscribed actions. Each factor has been theoretically pro-
posed to affect the experience of anger. First, the Type of Cause dimen-
sion is evident in major theoretical treatises on emotion (Averill, 1982;
Mandler, 1975, 1980). Second, anger that is perceived as due to the
mistreatment of people might be more frequent, intense, and enduring
than anger perceived as related to objects; this Focus distinction has also
been identi® ed in past theorising and research (Averill , 1982; Gates, 1926;
Scherer et al., 1986; Stein, Liwag, & Wade, 1994; Strayer, 1986) . Ander-
son (1991) also found the degree to which a cause was perceived as
interpersonal to be an important component in a layperson’ s causal analy-
sis. Third, Temporal Speci® city of an anger event is important because
ongoing , stable patterns of anger-eliciting behaviour may lead to more
intense or enduring anger than one-time, unique events (Weiner, 1982).
Additionall y, Anderson (1991) found that the dimension of `̀ temporary vs.
permanent’ ’ was signi ® cant in laypersons’ perceptions of causes. More-
over, each of these three factors can be found in some of the previously
cited coding schemes: Type of Cause emerged in all four codes; Focus is
represented in the Scherer et al. coding scheme; and two types of Temporal
Speci® city appear in Fisher and Johnson’ s scheme.
Study 2 was designed to determine the degree to which laypersons’
schemes of anger categorisation, as ascertained in Study 1, matched each
of ® ve psychological coding schemes. To determine these correspon-
dences, independent coders classi® ed each of the 60 descriptions provided
by mothers into one category for each psychological coding scheme.
Method
The clusters of anger causes generated by participants in Study 1 com-
prised the layperson classi ® cation schemes. To determine the ways that
each of the ® ve theoretical coding schemes would categorise the same
causes of anger, the transcripts of the 60 mothers’ descriptions used in
Study 1, complete with information about the type of relationship involved
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 641
(e.g. mother’ s anger at father, child’ s anger at mother, etc.), were given to
® ve coders (one of the authors, one graduate student, and three senior
psychology students). Each coder categorised all of the mothers’
responses with respect to each of the ® ve coding schemes. Coders exam-
ined all 60 responses within a relationship and all relationships (in random
order) within one coding scheme before moving on to another scheme
(provided in random order, except for the Carpenter and Halberstadt
scheme (submitted) which was given last). A response was designated as
`̀ codable’ ’ and was considered to be categorised if three of the ® ve coders
agreed on the category that best applied to the response.4
Results
For each type of relationship (parents to children, children to parents, and
parents to parents) and for each of the ® ve psychological coding schemes,
the amount of agreement between laypersons’ codings and psychologists ’
codings of the causes of anger was ascertained using the lambda statistic
(Goodman & Kruskal, 1954). Lambda is a measure of association appro-
priate for comparison of two classi ® cation systems consisting of nominal
categories (which need not have one-to-one correspondence; rectangular
matrices are acceptable). It ranges from 0.0 (when one classi ® cation does
not reduce prediction errors in the other classi ® cation) to 1.0 (when one
classi® cation perfectly predicts the second classi ® cation). Lambda has
recently been used by Haslam and Fiske (1992) to compare laypersons’
and psychological theorists’ classi ® cations of types of relationships; the
highest lambda obtained in their study was 0.33. In the present study, the
lambda statistic will represent the degree to which categories of each of the
® ve theoretical coding schemes correspond to categories derived from
students’ sorts of the same causes.
As can be seen in Fig. 1, the degree of overlap between layperson and
theoretical coding schemes varied by type of family relationship. Overall,
psychologists’ coding schemes showed more overlap with layperson clas-
si® cations of children to parent anger (mean lambdas = 0.36 for father and
0.33 for mother) than for the classi ® cations of parent to children anger
(mean lambdas = 0.18 for mother and 0.15 for father). In addition, for these
relationships , the degree of overlap was relatively consistent across psy-
chologis ts’ coding schemes. These patterns could indicate either that
conceptions of anger are clearer for children than for adults or that causes
4Detailed descriptions of how each of the ® ve psychological coding schemes categorised
the causes of anger, and comparisons of the relative merits of each coding scheme for
categorising the causes of anger in familial relationships , can be found in Carpenter and
Halberstadt (submitted). In that study, 86.7% of all categories were codable .
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
642 CAR PENTER AN D HALBERSTA DT
of parents’ anger at children are more complex or may be interpreted in
multiple ways. Codings of mother to father anger also showed good
average overlap (lambda = 0.35), but greater variability across psycholo-
gists’ coding schemes (lambda range = 0.19± 0.62) . The Carpenter and
Halberstadt scheme seemed to `̀ capture’ ’ best laypersons’ conceptions of
these anger causes. Thus, the psychologi cal coding schemes showed
differential correspondence to laypersons’ classi ® cations as a function of
type of family relationship, and for mother to father anger as a function of
psychologist coding scheme.
We also examined which psychologists ’ coding schemes showed the
most correspondence with laypersons’ classi ® cations. For mother’ s anger
at father, the Carpenter and Halberstadt coding scheme showed substantial
correspondence to laypersons’ organisation (lambda = 0.62). Other coding
schemes showed more moderate correspondence. For father’ s anger at
mother, four coding schemes showed similar overlap with laypersons’
sorts (lambdas = 0.29 for both Scherer et al., and Fisher & Johnson;
lambdas = 0.28 for both Camras & Allison, and Carpenter & Halber-
stadt). For parent to child anger, the Fisher and Johnson coding scheme
best coincided with laypersons’ classi® cations of anger causes (lambdas =
0.35 and 0.43 for mother to child and father to child anger, respectively).
FIG . 1 Degree of overlap, as measured by lambda, between 5 psychological schemes for coding
causes of anger as a function of type of relationship .
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 643
For child to parent anger, the Fisher and Johnson coding scheme was most
similar to laypersons’ conceptions of child’ s anger at mother (lambda =
0.38), but not father (lambda = 0.00). The Carpenter and Halberstadt and
the Scherer et al. coding schemes best matched laypersons’ sorts for child
to father anger (lambdas = 0.46 and 0.48) . These results highlight the
importance of distinguishing the type of relationship, and suggest which
psychologists’ coding schemes may be most correspondent with layperson
classi® cations.
The lambda statistic provides information about the degree of overall
association; however, it does not indicate where the overlaps in categorisa-
tions occurred. Therefore, to determine which layperson classi ® cations had
considerable correspondence with speci® c psychologists ’ categories, we
examined the ® ve frequency matrices that were created when layperson’ s
classi® cations were the rows and the psychologists’ codes were the col-
umns. For these descriptive purposes, `̀ considerable’ ’ category correspon-
dence was identi ® ed on the basis of two rules: (a) Rule 1: if 25% or more of
the items appeared in a single cell of the frequency matrix, and (b) Rule 2:
if a cell entry equalled or exceeded 5 items and this exceeded the expected
cell frequency by a factor of 3. Both rules are described because they
yielded different patterns of overlap between laypersons’ and psycholo-
gists’ categories.5
The rules were designed to be generous, such that
correspondences would be easily detected. Given the small frequencies
chosen to designate co-occurrences, the reported results should be treated
as tentative. The rules allowed all but 2 of the 15 layperson categories to be
considered for overlap with psychological coding schemes (these two
layperson categories consisted of fewer than ® ve items). These correspon-
dences can be seen in Table 4.
For mother’ s anger at father, four laypersons’ classi ® cations showed
substantial overlap with seven psychologists’ codes. For father’ s anger at
mother, three layperson classi ® cations overlapped with seven psycholo-
gists’ codes. Students and psychologists agree that parents’ anger at each
other is most likely due to violations of expectations compared to goal
blockages.
For both parents’ anger at children, one layperson classi® cation over-
lapped with one category in each of the ® ve psychologists ’ coding
schemes. Also, for only mother’ s anger at children, two layperson classi-
® cations overlapped with three psychologists’ codes. Both students and
psychologists agree that parents’ anger at children for disobedience and
5Rule 1 tended to identify overlaps at a more psychological level (e.g. for parents’ anger
at parents, Personality De® ciencies with Violation of an Expectation), whereas Rule 2
targeted co-occurrences at a more speci® c , pragmatic level (e.g. for children’ s anger at
parents, Inadequate Attention with Inappropriate Rewards).
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
644
TABLE 4
Corresponden ces Betw een Layp ersons’ and Psycholo gists’ C ategorisat ions of Anger
Causes for Each Rela tionsh ip
Laypersons’ Classi® cations Coding Rule and Psychologists’ Codes
Both Parents’ Anger at Each Other
Children Issues 2. Separation (Scherer et al.)
Money Issues 2. Lack of consensus (Fisher & Johnson)
Mothers’ Anger at Father
Money Issues 2. Failure to Reach Goals (Scherer et al.)
Not Being a Caring Partner 1. Ongoing Person-focused Expectancy Violations
(Carpenter & Halberstadt: C&H)
Not Contributing to Household 1. Violation of an Expectation (Shaver et al.)
2. Reversal or Loss of Power (Shaver et al.)
2. Ongoing Object-focused Expectancy Violation
(C&H)
Fathers’ Anger at Mother
Money Issues 2. Ongoing Object-focused Goal Blockage (C&H)
Personality De® ciencies 1. Failure to Conform to Social Norms (Scherer
et al.)
1. Violation of an Expectation (Shaver et al.)
1. Social Violations (Camras & Allison)
1. Not Ful® lling Role Expectations (Fisher &
Johnson)
Both Parents’ Anger at Children
Disobedience/Not Listening 1. Failure to Conform to Social Norms (Scherer
et al.)
1. Violation of an Expectation (Shaver et al.)
1. Social Violations (Camras & Allison)
1. Not Ful® lling Role Expectations (Fisher &
Johnson)
1. Ongoing Person-focused Expectancy Violations
(C&H)
Mothers’ Anger at Children
Disobedience/Not Listening 2. Not Helping (Fisher & Johnson)
Personality De® ciencies 2. Not Ful® lling Social Obligations (Fisher &
Johnson)
2. Speci ® c Person-focused Expectancy Violations
(C&H)
Children’ s Anger at Both Parents
Not Getting One’ s Way 1. Failure to Reach Goals (Scherer et al.)
1. Frustration of a Goal (Shaver et al.)
1. Not Receiving Desired Object (Camras &
Allison)
1. Not Helping (Fisher & Johnson)
Inadequate Attention 2. Inappropriate Rewards (Scherer et al.)
2. Justi ® able Damage (Camras & Allison)
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 645
personality de ® c iencies are most likely due to violations involving social
norms and expectations.
For children’ s anger at both parents, two layperson classi® cations
showed substantial overlap with six psychologists’ codes. For children’ s
anger at mother only, two layperson classi® cations overlapped with three
psychologists’ codes. For children’ s anger at father only, one layperson
classi® cation overlapped with one psychologis t code. The consensus of
both students and psychologists seems to be that children’ s anger at parents
is primarily due to obstacles to goals (mostly object-focused), but that
children also may be angered because of emotional expectancy violations.
Discussion
Study 2 was designed to determine the degree of correspondence between
laypersons’ and psychologists’ categorisations of the causes of anger and to
specify those correspondences. The degree of correspondence differed as a
function of (a) the particular psychologica l coding scheme being used, and
(b) the type of relationship being examined, as might be expected, given
that laypersons’ classi® cations differed across relationship types.
How well did the layperson and psychologis t categories correspond
overall? Of the eight layperson categories associated with parents’ anger
at each other, ® ve (Not Being a Caring Partner, Not Contributing to
Household, Personality De® ciencies, Money Issues, Children Issues) coin-
cided with at least one category for one parent in all ® ve psychological
categorisations . Of the ® ve layperson categories relevant to parents’ anger
at children, one (Disobedience/Not Listening) showed overlap for both
mothers and fathers with all ® ve of the psychologi cal coding schemes,
and for mothers only with one coding scheme. Another layperson classi-
® cation relevant only to mothers’ anger at children (Personality Deficien-
cies) corresponded with two psychologists ’ coding schemes. Finally , of the
® ve categories generated by laypersons to classify children’ s anger at
Children’ s Anger at Mother Only
Not Getting One’ s Way 1. Ongoing Object-focused Goal Blockage (C&H)
Unfair Treatment 2. Violation of an Expectation (Shaver et al.)
2. Not Ful® lling Role Expectations (Fisher &
Johnson)
Children’ s Anger at Father Only
Not Getting One’ s Way 1. Speci ® c Object-focused Goal Blockage (C&H)
Note : Rule 1 indicates that 25% or more of the items appeared in a single cell of the
frequency matrix. Rule 2 indicates that a cell entry equaled or exceeded 5 items and this
exceeded the expected cell frequency by a factor of 3.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
646 CAR PENTER AN D HALBERSTA DT
parents, one (Not Getting One’ s Way) overlapped for both parents with
each of the ® ve psychologi cal coding schemes. For children’ s anger at only
mother or only father, two other layperson classi ® cations overlapped with
four psychologists’ codes. Thus, approximately half of the layperson
classi® cations were also found in one or more of the psychologists’ coding
schemes.
In general, the overlaps between laypersons’ and psychologists’ cate-
gories characterised adult anger at both spouses and their children as being
due to expectancy violations , and children’ s anger as due to goal
blockages. This is the same pattern found in comparisons between psychol-
ogists ’ coding schemes (Carpenter & Halberstadt, submitted). This replica-
tion indicates the robustness of the phenomenon.
Layperson classi ® cations that were not represented in psychologists ’
coding schemes were as follows: for parents’ anger at each other, Inade-
quate Time and Being Controlling ; for parents’ anger at children, Naughty
Behaviours, Damaging Behaviours, and Interpersonal Harm; and for chil-
dren’ s anger at parents, Discipline and Power Struggles: 5 of the 10
nonoverlapping categories (some occurred for both mothers and fathers)
contained fewer than 10 items (see Tables 1 through 3), so that achieving
overlap as de ® ned by either rule would have been dif ® cult. However, the
remaining ® ve layperson classi ® cations seem not to correspond to any
categories of the ® ve psychological coding schemes. These results are
similar in magnitude to those of Carpenter and Trentham (submitted) who
found moderate correspondence when layperson and theorist classi ® cations
of subtype descriptions of females and males were made ignoring `̀ target’ ’
(female, male) boundari es. Therefore, the contention of Averill (1983) and
Anderson (1991) that laypersons’ conceptions of causes differ from those
of psychologists seems to be well founded. The position that layperson
knowledge may be unsystematic, however, is not supported. Students in
Study 1 were able to create meaningful clusters of anger causes.
We now need to ask what underlies the differences in psychologists ’ and
laypersons’ classi ® cations of anger causes. Layperson categorisations were
restricted by the particular anger descriptions availabl e for sorting, whereas
some psychologists ’ coding schemes were based on a wider array of anger
causes, relevant to a greater variety of relationships . Thus, some of the
psychologists’ categories may have been irrelevant to this particular data
set. However, two of the coding schemes (Carpenter & Halberstadt and
Fisher & Johnson) were developed with the express purpose of analysing
anger in families; these schemes did not do a better job of matching
layperson schemes than those not so designed. Thus, it is unlikely that
differences in layperson and psychologist classi ® cations are solely due to
differences in the descriptors being considered while categorisation
schemes were being developed.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 647
We suspect, instead, that laypersons and psychologists engage in differ-
ent types of analyses in their taxonomies of the causes of anger. Thus, the
next task is to identify the differences and similarities in the ways layper-
sons and psychologis ts think. Anderson (1991) distinguishes between
scienti ® c and phenomenal (i.e. experiential) accounts as well as between
explanatory and descriptive accounts of causes. We hypothesise that these
dimensions will distinguish laypersons from psychologists ; whereas psy-
chologis ts are typically `̀ doing science’ ’ (e.g. systematically explaining,
creating theories, establishing generic rules, etc.), laypersons are more
typically `̀ doing laundry ’ ’ (e.g. experiencing life and achieving everyday
goals) . Additionally , because prototypes of emotion vary along a `̀ ver-
tical’ ’ , hierarchical dimension (Shaver et al. 1987), and because other
research suggests that laypersons cluster at a more speci® c level whereas
psychologists operate at more abstract levels (Carpenter & Trentham,
1994) , we predicted that laypersons’ categories would re¯ ect more basic
levels of classi ® cation and psychologists ’ categories would re¯ ect more
superordinate inclusive levels. Study 3 was designed to determine the
degree to which psychologists ’ and laypersons’ categorisations differ
along these three dimensions: description vs. explanation, experiential
vs. scienti ® c, and speci® c vs. abstract.
STUDY 3
In Study 1, mothers’ descriptions of anger causes were grouped on the
basis of similarity , with no category names provided for making the
classi® cations. In Study 2, mothers’ descriptions were categorised on the
basis of pre-existing psychologis t coding schemes, so category labels had
been provided by those schemes. Thus, for Study 3, two units of analysi s
are appropriate: the category labels used in Study 2, and the unnamed
layperson grouping s created in Study 1. Therefore, in Study 3, participants
rated both the category labels and the grouping s without labels along the
three dimensions of: description vs. explanation, experiential vs. scienti® c,
and speci® c vs. abstract.
In Phase I of the study, students generated category labels for the
layperson grouping s of anger causes obtained in Study 1. This allowed
us to compare the category labels from psychologist coding schemes (see
Study 2) with the category labels we created for the layperson groupings
(see Study 1) and the category labels the laypersons created for the
layperson grouping s (Study 3). In Phase II of the study, an independent
set of participants rated: (a) the unnamed layperson groupings, and (b) the
three types of category labels just described, both with respect to the three
possible dimensions of variability.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
648 CAR PENTER AN D HALBERSTA DT
Method
Phase I: Generating Category Labels for LaypersonGroupings
Participants. Twelve undergraduate students generated category
labels for the layperson groupings . All four male students and four of the
female students had no children. However, four of the students were
mothers of children between the ages of ® ve and seven (similar to the
mothers who had generated the original anger descriptions) . One mother
was African-American; all other participants were Caucasian.
Materials. The descriptions of anger causes in each of the 15 layper-
son grouping s were compiled for each grouping separately. The groupings
were randomly ordered, then numbered.
Procedure. First, participants individuall y generated category labels
(`̀ names’ ’ ) for each layperson grouping . Each label was written on an
index card. These cards were collected and organised so that the 12 labels
(one for each partic ipant) for a single grouping were organised together.
Second, particants examined all 12 labels and independently ranked the 6
`̀ best’ ’ labels from this set. The labels most frequently ranked as ® rst by
participants were used as layperson category labels in Phase II.
Phase II: Dimension Ratings of Category Labels and ofGroupings
Participants. Five psychology majors in their senior year (three female
and two male) and six female psychology graduate students were invited to
participate. All students had substantial training in research methodology
and cognitive psychology. One undergraduate and one graduate student
were mothers. All participants were Caucasian.
Materials. For the category label-rating task, a list of 54 category
labels (4 for Camras & Allison; 6 for Carpenter & Halberstadt; 6 for
Fisher & Johnson; 5 for Scherer et al.; 5 for Shaver et al.; 15 for the
Carpenter & Halberstadt labels of layperson grouping s, and 13 for the
layperson labels of layperson groupings from Phase I was prepared.
Descriptions of the dimensions to be rated were presented on separate
pages, followed by the 54 labels. The three dimensions were presented
in counterbalanced order across participants. Thus, participants rated all 54
labels for one dimension prior to going on to the next dimension.
For the grouping-rating task, the data set was reduced to make the task
feasible for the raters and to focus on the speci® c question of similarities
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 649
and differences between laypersons’ and psychologists’ cognitive repre-
sentations of anger causes. First, three of the six relationship types were
chosen, such that (a) one relationship showed maximum overlap between
psychologists’ and laypersons’ categorisations : mother to father (lambda =
0.35), and (b) two relationships showed minimum overlap: mother to child
(lambda = 0.18) , and child to mother (lambda = 0.15). This `̀ degree of
overlap’ ’ variabl e allowed us to determine the extent to which the three
dimensions being examined contribute to the differences previously
obtained between laypersons’ and psychologists’ categorisations . We
expected that categorisations of the mother to father relationship would
be rated similarly across the three dimensions, but that the mother to child
and the child to mother anger categorisations would be rated differently on
at least one of the three dimensions. Second, three of the psychological
coding schemes were chosen to be rated, such that, for most of the
relationships being examined, the greatest and least overlap between
laypersons’ and psychologists ’ categorisations were represented. The
empirical pattern of overlaps suggested that the Camras and Allison,
Fisher and Johnson, and Carpenter and Halberstadt coding schemes might
yield the most useful data on similarities and differences. Fortunately, from
a theoretical perspective, all three coding schemes were speci® cally
designed to provide taxonomies for causes of anger in the family. Follow-
ing this reduction, 57 grouping s remained to be rated.6
For both the label-rating task and the grouping-rating task, the defini-
tions of the three dimensions were the same. As an example, the experi-
ential (rating = 1) versus scienti ® c (rating = 7) distinction was described as
follow s:
Experiential accounts of behaviour tend to be more subjective, indicating
how people `̀ experience’ ’ an event. These accounts refer to one’ s `̀ stream of
consciousness’ ’ and may also include feelings associated with an event.
Scienti® c accounts tend to be more objective, indicating characteristics a
behaviour has in common for most people. Scienti ® c accounts therefore tend
to be more `̀ summary’ ’ in nature. For example, `̀ when my father teases me,
it hurts my feelings’ ’ is an experiential account. A scienti® c account of the
same event might be `̀ my father exhibits passive-aggressive behaviour
toward me’’ .
6The 57 groupings were composed as follows: 15 groupings for the layperson sorts, 14
groupings for the Carpenter and Halberstadt coding scheme, 17 groupings for the Fisher and
Johnson coding scheme, and 11 groupings for the Camras and Allison coding scheme. Across
these four schemes, there were 20 groupings for the mother to father relationship, 16
groupings for the mother to child relationship, and 21 groupings for the child to mother
relationship.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
650 CAR PENTER AN D HALBERSTA DT
The dimension of description (rating = 1) versus explanation (rating = 7)
distinguished between description of behaviour (e.g. `̀ She checked the
door locks three times’ ’ ) and an explanation of behaviour (e.g. `̀ she is
nervous about intruders ’ ’ ). The dimension of speci® c (rating = 1) versus
abstract (rating = 7) distinguished between speci® c/exclusive descriptions
(e.g. rocking chair, coffee table) and more generic/inclusive descriptions
(e.g. furniture).
Procedure. First, the intent and procedure of the study was explained
to participants. Then each participant was instructed in the use of the scales
and the de ® nitions of the dimensions until the experimenter was satis ® ed
that the participant understood the task. In all cases, partic ipants rated
category labels ® rst. Our rational e was that this task would acquaint
participants with the task and rating dimensions; we expected ratings of
the grouping s to be more dif ® cult. This expectation was supported by
informal debrie ® ngs of the participants, who described the label rating
task as `̀ dif® cult’ ’ and the grouping rating task as `̀ extremely dif ® cult’ ’ .
In addition, rating of the experiential /scienti® c dimension was described as
most dif ® cult. Participants completed both tasks at their own pace (70 ± 90
minutes).
Results
Reliabil ity of Ratings. We were concerned about the consistency of
both the label and grouping ratings because participants expressed having
dif ® culty with both tasks. Coef® cient alphas were therefore computed for
each dimension, across participants. For the label-rating tasks, alphas were
remarkably high for the speci® c versus abstract (r = 0.86) and the descrip-
tion versus explanation (r = 0.91) dimensions. Alpha for the experiential
versus scienti ® c dimension was lower (r = 0.70) than for the other dimen-
sions, as might be expected given participants ’ reports of this dimension
being the most dif® cult to rate. However, this reliability coef ® cient is quite
`̀ respectable’ ’ ; it is comparable to that obtained for widely used tests using
Likert scales (e.g. Edwards Personal Preference scales, rs = 0.60 ± 0.87;
Locus of Control , r = 0.70; both reported in Anastasi , 1988).
For the grouping ratings, reliabilities were computed for each dimension
in each of the three relationships, yielding nine alpha coef® cients. Eight of
these reliability estimates were surprisingl y good (ranging from 0.67 to
0.81), considering the dif ® culty of the task. The alpha for the grouping s of
mother anger toward father, rated in terms of abstraction, however, was
extremely low (r = 0.16) . Patterns obtained for this one relationship/
dimension combination should therefore be interpreted cautiously .
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 651
Label Ratings. These data were collected to determine whether the ® ve
sets of psychologists’ category labels differed from the two sets of layper-
son labels along the three dimensions of description versus explanation ,
experiential versus scienti® c, and speci® c versus abstract. For each dimen-
sion, each participant’ s average rating within each of the seven sets of
labels was determined (e.g. average rating for the four Camras & Allison
categories, etc.). Then the ratings for the ® ve psychologists labels were
compared to each other (in a 1-way ANOVA) and the ratings of the two
layperson labels were compared to each other (in a t-test). These two
analyses showed the psychologists ’ labels to be similar to each other
across all three rating dimensions (all Fs < 1), and the layperson labels
to be similar across all three dimensions (all P-values > 0.26). Thus,
psychologists’ and laypersons’ labels seem to be `̀ internally consistent’ ’
with respect to the dimensions of differentiation being tested. To determine
whether layperson and psychologis t labels differed, a t-test was conducted
for each rating dimension.
For the dimension of speci® c versus abstract, psychologists ’ labels were
rated as signi® cantly more abstract (M = 4.34, SD = 0.80) than labels of
layperson grouping s [(M = 3.75, SD = 0.63) ; t(52) = 9.07 , P = 0.004] . For
the dimension of description versus explanation, no signi ® cant difference
was obtained between psychologist (M = 3.79, SD = 0.69) and layperson
labe ls [ (M = 3.65, SD = 0.61) ; t < 1] . For the dimension of experiential
versus scienti ® c, the analysi s revealed psychologists ’ labels to be more
scienti ® c (M = 4.26, SD = 0.65) than layperson labels [ (M = 3.85, SD =
0.77 ); t(52) = 4.38, P = 0.04] . Thus, using category labels as the unit of
analysis, psychologists ’ and laypersons’ categorisations of anger causes
were shown to differ both in terms of the `̀ speci® city vs. abstraction’ ’ and
the `̀ experiential vs. scienti ® c ’ ’ dimensions.
Grouping Ratings. For each participant, ratings of grouping s were
averaged within each relationship and each rating dimension. For each of
the nine relationship/dimension combinations , t-tests compared ratings of
psychologist grouping s to those of layperson groupings. Due to the number
of comparisons, a Bonferroni correction was used to set the accepted
probability level at 0.0055.
For mother to father anger no differences were expected for the varia-
bility dimensions; this was the relationship that had shown greatest overlap
between psychologist and layperson grouping s in Study 2. This prediction
was supported. For each dimension, t-tests showed ratings of psycholo-
gists’ and laypersons’ grouping s not to differ: (a) speci® c/abstract, t = 0.05 ,
P = 0.96, (b) description/explanation, t = 0.85, P = 0.44, and (c) experi-
ential /scienti ® c, t = 0.12 , P = 0.91 .
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
652 CAR PENTER AN D HALBERSTA DT
Overlaps between layperson and psychologis t categorisations were
minimal for the mother to child and the child to mother relationships.
We therefore expected differences in ratings of grouping for one or more
of the three variability dimensions. For mother to child anger, participants
rated psychologis t grouping s as more abstract (M = 3.70, SD = 0.22) than
layperson groupings [ (M = 3.14, SD = 0.20), t = 5.76 , P = 0.004] . Ratings
for the other two dimensions were not different for psychologis t and
layperson groupings (both ts < 1). For child to mother anger, no significant
differences were obtained for any of the three dimensions: (a) speci® c/
abstract, t = 2.92, P = 0.03, (b) description/explanation, t < 1, and (c)
experiential/scienti ® c, t = 1.95, P = 0.11. In summary, when groupings are
the level of analysis, the dimension of speci® c versus abstract best differ-
entiated between psychologists’ and laypersons’ classi ® cations.
Discussion
Study 3 identi ® ed the dimension of speci® c/abstract as best differentiating
between psychologists’ and laypersons’ categorisations (for those relation-
ships that showed strong differentiation) in terms of both labels and
grouping s. The dimension of experiential/scienti ® c also distinguished
between those categorisations at the level of category labels. These two
variables may partiall y account for the lack of overlap between psycholo-
gists’ and laypersons’ categorisations obtained in Study 2. The dimension
of description/explanation was not shown to be implicated in the differ-
ences between the two types of categorisations . This variabl e may partiall y
account for the degree of overlap that was found between the two types of
categorisations . This makes sense, considering the stimuli and task. First,
items within groupings were descriptions of anger experiences or causes.
The psychologists ’ coding schemes we chose to study categorised causes of
anger, to a great extent, in terms of explanato ry frameworks. Similarly,
when asked to sort diverse descriptions of anger into categories, laypersons
may have been relying on explanatory similarity in order to complete the
task. Future research could compare results of `̀ free’ ’ sorts to those of
`̀ directed’ ’ sorts (requiring subjects to sort descriptions in terms of expla-
nations) .
Studies 2 and 3, in combination, address two research questions posed in
this paper: `̀ How do perceived anger causes differ across relationships ? ’ ’
and `̀ How do psychologists’ and laypersons’ categorisations of such anger
causes differ? ’ ’ Psychologists’ and laypersons’ categorisations are similar
in many respects, but differ reliably in terms of abstraction, and also
possibly in terms of the experiential/scienti® c dimension. However, the
type of relationship in question must also be considered. Relationship
affected both the degree of overlap between categorisations and the factors
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 653
that in¯ uence differences: (1) the mother to father relationship showed a
good deal of overlap for psychologists ’ and laypersons’ categorisations; (2)
the mother to child relationship showed poor overlap, partiall y due to level
of abstraction being used during categorisation; and (3) the child to mother
relationships showed poor overlap that was not due to differential levels of
abstraction being used. This differential pattern across relationships sup-
ports our contention that anger is socially constructed both by psycholo-
gists and laypersons; behaviours can be differently interpreted depending
on relationship context.
G ENERAL DISCU SSION
The goals of the three studies reported were (a) to generate a taxonomy of
laypersons’ perceptions of anger causes, so we could learn whether these
perceived anger causes (b) differ across types of relationships (in this case,
family relationships) , (c) are cognitively represented as prototypes, and (d)
correspond to psychologists ’ taxonomies of anger causes. In Study 1,
students demonstrated a clear and consistent pattern of categorising
descriptions of anger causes; 15 distinguishable categories were gener-
ated. These categories provide a taxonomy of perceived anger causes
within nuclear family relationships.
As predicted, college students clustered anger causes differently across
types of family relationships, except for the two categories of Inadequate
Time or Attention and Personality De® ciencies, which are experienced
across several relationships. Given the age and power differences within
the family structure, and the probable differences in expectations, goals,
and activities as a consequence of those age and power differences, the
differences in anger causes are not surprising . They are, however, very
interesting from a descriptive perspective, and they additionall y highlight
the importance of examining emotional experience within the context of
social relationships.
With respect to cognitive representations , students’ structuring of anger
causes matched the characteristics of a prototype. That is, several cate-
gories of anger causes within each type of familial relationship were
elicited, as expected in a prototypic representation. In addition, these
categories varied across relationships and levels of speci® city (e.g. money
issues vs. personality de ® ciencies). Finally , as expected in a prototypical
structure, the category boundari es were sometimes fuzzy; some items `̀ ® t’ ’
into more than one category. Such data suggest that a taxonomy of the
causes of anger (as well as other emotions) is possible for both descriptive
and predictive purposes.
It is interesting to note the degree of similarity that this taxonomy shares
with another taxonomy of causes created for a very different content
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
654 CAR PENTER AN D HALBERSTA DT
domain. In students’ phenomenal organisation of the causes of success and
failure, the categories of personality traits and control concerns emerged
(as well as several other categories of causes; Anderson, 1991). These two
categories of success and failure causes parallel the two categories of
Personality De® ciencies and Disobedience/Not Listening of anger causes
derived from our participants’ sorts in Study 1. Thus, how people think
about the causes of anger may be similar to how people think about the
causes of success and failure. It may be that speci® c types of causes (e.g. of
anger, success, etc.) are connected, at an even more superordinate hier-
archical level, to a more generic prototype of cause. Additionally , it may be
possible to create taxonomies about phenomenologi cal components to
summarise similarities (and differences) in people’ s thought processes
across very different content domains.
Finally , comparison of the students’ categorisation schemes to ® ve
psychological coding schemes in Study 2 resulted in both similarities
and dissimilarities. First, layperson classi ® cations appeared to be much
more pragmatically structured (i.e. focused more on topic matters such
as money issues or discipline), whereas the psychologi cal codes focused
more on underlying emotional experience (such as violation of a personal
or social contract or loss of power, status, respect). Second, only half of
laypersons’ classi ® cations were found to coincide with any categories of
psychologists’ coding schemes. Study 3 provided evidence that some of the
differences were due to the level of abstraction being used for classifying
causes. The presence of this `̀ level of abstraction’ ’ dimension offers
additional support for our contention that causes of anger are represented
as prototypes. However, this dimension was not shown to differentiate
between causes across all relationsips ; this con® rms our expectation that
causes of anger need to be studied with respect to particular contexts, as the
social constructionist position would dictate.
In summary, a taxonomy of anger causes was generated by laypersons,
and this taxonomy is highly sensitive to familial relationship differences, is
structured as a prototype, and shows both differences and similarities to
psychologists’ taxonomies of anger causes. Given the differences, which
taxonomy of anger is most useful for understanding anger and its causes?
This decision must obviousl y be based on the goals of the research. For
example, researchers concerned about predicting people’ s behaviours sub-
sequent to an anger episode might prefer to use laypersons’ classi® cations,
because the causal attributions of the person being angered may profoundly
affect the phenomenal experience of that anger. Liu, Karasawa, and Weiner
(1992) maintain that negative emotions (such as anger) give rise to causal
analyti c reasoning and to more explicit attribution searches than do posi-
tive emotions. According to their research, an angry person is likely to
attribute a negative emotion to a single, suf® cient cause. Such an attribu-
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
W HAT M AKES PEO PLE AN G RY ? 655
tional analysi s would allow for a clear direction of action in response to the
anger. Therefore, people’ s own perceptions of the reasons they became
angry may be better predictors of subsequent behaviours than are psychol-
ogists ’ theories. However, understanding the underlying factors (e.g. sense
of being unjustly treated, loss of respect) rather than the ostensible causes
(e.g. discipline , children’ s issues) in their relationships may help indivi -
duals to initiate responses that will optimise outcomes. Thus, clinicians
who wish to test clients’ understanding of each others’ underlying anger
causes may prefer to use one or more of the psychological coding schemes.
Also, researchers may wish to test the ef® cacy of different behavioural
responses to anger causes when categorised according to their underlying
structures. The present research now makes availabl e a layperson taxon-
omy organised for anger, provides a comparison of taxonomies conceived
both by several psychologi cal coding schemes and by laypersons, docu-
ments their relative merits, and indicates the correspondence between these
schemes across a variety of family relationships .
Manuscript received 6 July 1994
Revised manuscript received 26 February 1996
REFEREN CES
Anastasi, A. (1988). Psychological testing (6th ed.). New York: Macmillan.
Anderson, C.A. (1991). How people think about causes: Examination of the typical phe-
nomenal organization of attributions for success and failure. Social Cognition, 9, 295±
329.
Aries, E. (1987). Gender and communication. In P. Shaver & C. Hendrick (Eds.), Sex and
gender (pp. 149± 176). Newbury Park, CA: Sage.
Averill , J.R. (1982). Anger and aggression: An essay on emotion. New York: Springer.
Averill , J.R. (1983). Studies on anger and aggession: Implication for theories of emotion.
American Psychologist, 38, 1145± 1160.
Bengtson, V.L., & Black, K.D. (1973). Intergenerational relations and continuiti es in
socialization. In P.B. Baltes & K.W. Schaie (Eds.), Life-span developmental psychol-
ogy: Personality and socialization (pp. 207± 234). New York: Academic Press.
Brewer, M.B., & Lui, L.N. (1989). The primacy of age and sex in the structure of person
categories. Social Cognition, 7, 262± 274.
Camras, L.A., & Allison, K. (1989). Children’ s and adults ’ beliefs about emotion elicita-
tion. Motivation and Emotion, 13 , 53 ± 69.
Carpenter, S., & Halberstadt, A.B. (submitted). Causes of anger in family relationships.
Carpenter, S., & Trentham, S. (submitted). Subtypes of females and males: A categorical
analysis.
Fehr, B., Russell, J.A., & Ward, L.M. (1982). Prototypicality of emotions: A reaction time
study. Bulletin of the Psychonomic Society, 20, 253± 254.
Fisher, C.B., & Johnson, B.L. (1990). Getting mad at mom and dad: Children’ s changing
views of family con¯ ict. International Journal of Behavioral Development, 13, 31± 48.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014
656 CAR PENTER AN D HALBERSTA DT
Fisher, C.B., Reid, J.D., & Melendez, M. (1989). Con¯ ict in families and friendships of
later life. Family Relations, 38, 83± 89.
Gates, G.S. (1926). An observational study of anger. Journal of Experimental Psychology,
9, 325± 336.
Goodman, L.A., & Kruskal, W.H. (1954). Measures of association for cross classi® cations.
Journal of the American Statistical Association, 49, 732± 764.
Halberstadt, A.G., & Carpenter, S. (1994). How mad are you? The frequency, intensity,
duration, and onset latency of anger in family relationships. Unpublished raw data.
Halberstadt, A.G., Fox, N.A., & Jones, N.A. (1993). Do expressive mothers have expressive
children? The role of socialization in children’ s affect expression. Social Development, 2,
48± 65.
Haslam, N., & Fiske, A.P. (1992). Implicit relationship prototypes: Investigating ® ve
theories of the cognitive organization of social relationships . Journal of Experimental
Social Psychology, 28, 441± 474.
Liu, J.H., Karasawa, K., & Weiner, B. (1992). Inferences about the causes of positive and
negative emotions. Personality and Social Psychology Bulletin, 18, 603± 615.
Maltz, D.N., & Borker, R.A. (1982). A cultural approach to male ± female miscommunica-
tion. In J.J. Gumperz (Ed.), Language and social identity (pp. 195 ± 216). Cambridge:
Cambridge University Press.
Mandler, G. (1975). Mind and emotion. New York: Wiley.
Mandler, G. (1980). The generation of emotion: A psychological theory. In R. Plutchik &
H. Kellerman (Eds.), Emotion: Theory, research, and experience (pp. 219± 243). New
York: Academic Press.
Miller, G.A. (1969). A psychological method to investigate verbal concepts. Journal of
Mathematical Psychology, 6, 169 ± 191.
Russell, J.A. (1991). In defense of a prototype approach to emotion concepts. Journal of
Personality and Social Psychology, 60, 37± 47.
Savitsky, J.C., Czyzewski, D., Dubord, D., & Kaminsky, S. (1975). Age and emotion of an
offender as determinants of adult punitive reactions. Journal of Personality, 44, 311± 320.
Scherer, K.R. (1984). Emotion as a multicomponent process: A model and some cross-
cultural data. In P. Shaver (Ed.), Review of personality and social psychology (Vol. 5, pp.
37± 63). Beverly Hills, CA: Sage.
Scherer, K.R., & Tannenbaum, P.H. (1986). Emotional experiences in everyday life: A
survey approach. Motivation and Emotion, 10 , 295± 314.
Scherer, K.R., Wallbott , H.J., & Summer® eld, A.B. (1986). Experiencing emotion: A cross-
cultural study. New York: Cambridge University Press.
Shaver, P., Schwartz, J., Kerson, D., & O’ Connor, C. (1987). Emotional knowledge:
Further exploration of a prototype approach. Journal of Personality and Social Psychol-
ogy, 52, 1061± 1086.
Strayer, J. (1986). Children’ s attributions regarding the situational determinants of emotion
in self and others. Developmental Psychology, 22, 649± 654.
Stein, N.L., Liwag, M.D., & Wade, E. (1994). A goal-based approach to memory for
emotional events: Implications for theories of understanding and socialization. In R.D.
Kavanaugh, B.Z. Glick, & S. Fein (Eds.), Emotion: The G. Stanley Hall Symposium.
Hillsdale, NJ: Lawrence Erlbaum Associates Inc.
Weiner, B. (1982). The emotional consequences of causal attributions . In M.S. Clark & S.T.
Fiske (Eds.), Affect and cognition: The seventeenth annual Carnegie symposium on
cognition (pp. 185± 209). Hillsdale, NJ: Lawrence Erlbaum Associates Inc.
Dow
nloa
ded
by [
UQ
Lib
rary
] at
18:
45 1
5 N
ovem
ber
2014