the domain of emotion words on lfrluk

CATHERINE LUTZ-State Universitv of New York, Binghamton

Internal feeling states have commonly been assumed to be the primary referents of emo-
tion words in Western thought, both social-scientific (Cardiner, Metcalf, and Beebe-Center
1937;Solomon 1976,1978)and lay (Davitz 1969).I n these views, the function of the emo-
tion word i s to label an internal state and perhaps to communicate that state to others. To
an important extent, however, scientific conceptions of emotion and person are rooted in
American ethnopsychological themes (LeVine 1980;Lutz 1981a)and everyday uses of emo-
tion language (Kagan 1978). Examination of the use of emotion words among several
Oceanic peoples, including Samoans (Cerber 1975). Pintupi Aborigines (Myers 1979),and
A’ara speakers of the Solomon Islands (White 1981).reveals an alternative view of emotion.
In these societies, emotion words are seen as statements about the relationship between a
person and an event (particularly those involving another person), rather than as statements
about introspection on one’s internal states. In addition, physiological descriptions of emo-
tion do not frequently occur naturally and in some cases are extremely difficult to elicit
(Cerber 1975:183).
A study of emotion words on the Micronesian atoll of lfaluk indicates that these words
are defined and sorted by informants based on the situations in which the emotion usually
occurs. The way in which people talk about emotion words i s related to broader
ethnotheories about the nature of the self. The world view and values within which emo-
tion words have meaning are therefore emphasized. The clusters and dimensions that
emerge from card sorting by lfalukian informants display both some universal features of
semantic meaning and some culturally specific foci that are related to the central values
and ethnotheories of many Ifaluk. Although the differences between emotion words and
emotions are clear, both cultural differences and universals in the meaning of emotion
words can provide important evidence about the nature of emotion itself (Lutz 1981 b).

ethnographic background and method

The lfaluk are a Malayo-Polynesian group inhabiting a one-half square mile coral atoll in
the Western Pacific.’ The 430 members of this densely populated society subsist on taro,

The cognitive organization of the domain of emotion words on the island of

lfaluk is examined. Native speakers define and sort emotion words based on the
situation in which the relevant emotion typically occurs. English emotion words,
bv contrast, are usuallv organized on phvsiological principles. Universal dimen-
sions of meaning found in other studies of psvchological language and culture
are also found here, however, and these dimensions, being themselves
multidimensional, are interpreted in light of the lfalukian ethos. [cognitive an-
thropology, emotion, ethnopsychology, psychological anthropology, Ifaluk]

Copyright @ 1982 by the American Ethnological Society

009404%/82/010113-1652.1 O/l
 breadfruit, coconut, and fish. Hereditary chiefs head the several matrilineal clans and meet
periodically to decide on island-wide issues, but they are accorded few special privileges.
Extended matrilocal households average 13 members each and, as adoption is quite com-
mon, consist of members of several clans and of varying degrees of biological relatedness.
The most striking features of lfaluk values are their emphasis on nonaggression, on
cooperation and sharing within and between households, and on obedience within a system
of ranking. Murder i s unknown; the most serious incident of aggression in a year involved
the touching of one man’s shoulder by another, a violation that resulted in the immediate
payment of a severe fine. People are expected to share their labor, food, and goods, and
those who shirk such responsibilities are sharply criticized. Communal labor and food shar-
ing are daily occurrences, and the value of sharing one’s children i s expressed in an adop-
tion rate of over 40 percent. Finally, the system of rank hierarchically orders clans,
lineages, and individuals. The main prerogative of rank i s that of exacting obedience from
others. The status of rank carries with it, however, the obligation to speak and act “kindly”
to those of lower rank.
These values and their behavioral expressions are intimately linked to the environmental
pressures and constraints of an island that is small, bounded, and subject to periodic
devastation by typhoons. In such a context, cooperation, obedience to authority, sharing of
resources, and nonaggression may contribute to survival. The cultural order that is built
against this backdrop will be seen below to be reflected in the meaning and cognitive
ordering of emotion words.
The methods by which the domain of emotion words was investigated will now be
reviewed. Although no word in the lfalukian language translates as “emotion,” 31 words
that were identified as “about our insides” were selected for use in the formal testing ac-
cording to the following procedure. From the beginning of the fieldwork period, all terms
were collected that could conceivably have relation to emotion, thought, personality, or
expressive behavior. These words came from (1) a search of Sohn and Tawerilmang’s (1976)
dictionary of the virtually identical dialect of the neighboring island of Woleai; (2) daily
conversations in which states or traits of individuals were discussed; and (3) a collection of
words “relating to emotion” from a bilingual informant.’
Interviews based on the resulting l i s t of 208 words revealed several classes of words, in-
cluding wegitegil (behavior) and ununul varemat (personal style). These words were said
“not [to be] about our insides” (tai kofal niferash). An informant was asked to sort the 208
words into three groups: ”behavior,” ”personal style,” and “all others.” The latter group
consisted of 58 words, the definitions of which were obtained from between three and ten
informants of both sexes. These words, written on small cards, were free-sorted into piles
by six literate informants. The sorters were told that the piles could be of any size and that
words should be sorted according to whether they were similar (hafitegl or a little similar.
Any two words that were sorted into one pile by all six people were considered
synonymous. Several people were asked to identify the more commonly used word of the
pair, and the other was then omitted from further testing. Based on the definitions, archaic
or nonnative words were also eliminated.
This process resulted in a l i s t of 31 words that were then sorted by 1 3 informants into
piles according to the procedure described above. The order of, and changes in, sorting of
each word were noted, and when the task was complete each individual was asked why all
of the words in each group went together.)
The process just outlined constituted an attempt to elicit the domain and classification
of emotion words in the absence of an overarching category of emotion, as distinct from
thought or will. Although there i s not one word like “emotion” that exclusively defines the

114 amwican ethnologist

l i s t that resulted, all of the words are ”about our insides.” While some related terms may
never have been discovered, informants agreed on the cohesiveness of the group that con-
stituted the list.

emotion word clusters and the meanings of situations

The groupings made by informants were analyzed both for structure (which will be ex-
amined first) and for underlying dimensions of meaning. A ”hierarchical clustering” pro-
gram was used that bases its calculations on a similarity matrix consisting of the number of
times each word was paired with each of the others by informants.‘ The criteria used to
combine the similarity measure between each pair of emotion words and each subsequent
cluster were the average measures of similarity between the clusters. The diagram that this
program produced may be conceptualized as the “average sorting” performed by the
lfaluk and i s presented in Figure 1. There are five major clusters in this domain and they will
be interpreted below in view of the explanations that those who did the sorting gave for
their inclusion of words in a single pile.
The most frequent criterion on which informants sorted and defined each emotion word
was situation. Typical explanations for the sorting of several words into the same pile in-
cluded, “If my child dies, I will feel all of them,” or “If there i s gossip about me in the
village, I’ll feel ,- and then I’ll feel -.I’ Two words were grouped together if
(1) the emotions are felt in the same or similar situations, or (2) the emotions may follow
one after the other in a hypothetical sequence of situations. An example that includes both
types of sorting rationale is the following explanation given for one four-word pile:

If someone goes away on a trip, you feel livernarn [longing] and lalornweiu [loneliness/sadness], and
if you had nothing t o give them [as a going away gift] you feel tang [frustration/grief] and filengaw
[incapable/uncomfortable].

Other bases besides situation were sometimes used for sorting explanations, and these will
be noted in the analysis of each cluster which follows.

cluster I: the emotions of good fortune This group of words is separate from all others
at the highest level of branching in Figure 1. These words were rarely sorted with others out-
side of this group. When situations were mentioned as the basis for sorting, they were often
situations in which “good” things happen. Definitions also reveal that all of these emotions
are pleasant ones that occur in the context of good fortune. The reasons that sorters gave
for including them in one pile included the following: “They all involve something happen-
ing that we tipeli [want/desire it]”; “They all mean ker [happiness/excitement]”; and a fre-
quent response, that they all are associated with having a boyfriend or a girlfriend. Many of
the emotions also occur when a person has a valuable object. This cluster includes several
words that are taboo in mixed company, as they refer specifically or especially to emotions
associated with sexual situations. No other cluster contains a taboo word.
In several cases, common behaviors that follow the emotions in this cluster were men-
tioned as one of the bases of sorting. A person i s these states walks about the village, is
talkative, and tends to act in a disruptive, “show-off,” or improper manner. Hence, the
emotions in this cluster are not morally good or the outcome of ethically correct behavior
in a certain situation. Rather, they simply occur in rewarding situations. These rewards are
not conceived of as internal or physiological, and in only one case was the internal state
itself elaborated as a basis of sorting-”You feel like you are great, the number one
person.”

emotion words on lfaluk 115

 CHEGAS (romantic pride)
8OSU(excite~nt/jcalousy~ -

II 8080 (disappointment)
METAGU (fear/onxiety)
RUS (panic /fright /surpr is 8 )
LUGUMET (discomfort/gu ilt)
MA (sharne/emborrassment )

LALOlLEffG (insecurity)

IE GASECHAULA (hate) D-
LIffGERJNGER (irritation)
TlPMCXMUXH (8 hor t- temper )
SONG (justified anger)

Defining Situations Underlying Major Branchings:

A - Pleasant VS. unpleasant rewards of the situation

B - Danwrous vs. nan-dangerous situation
C
D
-- Certain vs. u n c e r t a i n s i t u a t i o n
Misfortune vs. human error and conflict at base of situation

Figure 1. Hierarchical clustering diagram of 31 lfalukian emotion words (not to scale;

index of structure = ,941.

cluster 11: the emotions of danger This cluster contains words that all refer to
dangerous situations. Several people said that they linked metagu (fearlanxiety), lugumet
(discomfort/guilt), and ma (shamelembarrassment) because each i s felt when one must go
someplace where respectful behavior i s required. These situations include walking past a
seated group of elders or past one’s brother (if female), walking on the chief’s property, or
when one goes to visit someone but has nothing to give them, especially food. They are
also sorted together because they are felt when one walks up to a group of people. The
danger in the latter situation is due to the expectation that such people “will laugh at you.”
Bobo (disappointment) was most often sorted with rus (panic/fright/surprise) because
they both involve ”something bad happening that you don’t expect.” Metagu (fearlanxiety)
was the word that formed a bridge between rus and bobo, ma and lugumet in many
people’s sortings. It i s important to note that this second cluster was also quite distinct from
any other (see Figure 1).The perception of danger in a situation may be of a higher order of
importance than any other factor besides that of the pleasantness or unpleasantness of the
consequence of a situation.

C M M: i h -8 d c m t h arwl k r s This cluster i s a generally unified one;

that is, almost all individuals sorted the words in it together. There is also uniformity in the
reasons given for the sortings. Each of the four emotions-fago (compassion/Iove/sadness),
pak (homesickness), liyemam (longing), and lalomweiu (lone1iness/sadness)- is felt in the
context of separation or the threat of it. Adolescent boys who must go to the distant island
of Ulithi to attend high school were said to experience all four emotions; in these states,
“their thoughts are random” and “they think only of home.” Death also causes all four
emotions to be evoked and was mentioned as the basis for sorting.
In cases where laloileng (insecurity) was sorted with the other words in cluster 111, it was
due to their common evocation during sickness. According to one older man,
If I‘m sick, I feel both laloileng and lalomweiu that I will die and no one will get food and sweet co-
conut toddy for my children. I feel liyemam if a relative dies who used to always be in my house, and
it made me feel good to see him. But I feel liyemam because there’s no one to talk to me.
Although this cluster may be described as the “emotions of loss,” it also contains most of
the emotion words that refer to positive connections between individuals, even while those
connections may be in jeopardy due to death, travel, or sickness. The most commonly used
word in the cluster, fago, refers to one’s relationship with a more unfortunate other, rather
than to a more atomistic internal feeling such as “sadness.” Similarly, lalomweiu and
liyemam are felt by those who remain in the more fortunate position of being on the island
and healthy. Thus, the core words of this cluster focus both on attachment to a more unfor-
tunate other, as well as on the loss suffered by the self.

W : i h ----II--- . I This cluster contains most of the emotions

that are considered the most unpleasant to experience. The focus on this aspect of emotion
i s shown by the fact that here internal feeling tones were the basis for sorting more often
than in any other cluster. It was frequently said that these emotions belong together
because “they all feel bad inside.” Song (justified anger), the most commonly used term in
the cluster, i s considered “good”; however, this “goodness” is of a higher, moral level, as it
indicates that one has taken proper notice of the violation of a cultural norm by another.
The situations that each of these emotion words refer to include the element of human
error, and each word further distinguishes between whether that error i s one’s own or
another’s. Lingeringer (irritation/small anger) and tipmochmoch (short-temper) are milder
and less justifiable responses to frustration than song, as they refer to the thwarting of in-
dividual, rather than culturally defined, goals. Tang (frustrationlgrief) refers to the response
to situations where one i s thwarted either justly or unjustly, but in either case one i s
powerless to do anything about it. Thus, one person said that all of the words in cluster IV
belong together because they can all occur when one is interrupted in one’s work. Song will
be felt if this interruption i s uncalled for; lingeringer, tipmochmoch, and tang are all felt if
one is called away from work to perform an errand for someone of higher rank. As the
value d obedience to elders i s paramount, such a request i s acceptable and supercedes
other culturally constituted goals. Casechaula (hate) is also felt in the same situations, but
is said to follow upon the other feelings.
Less centrally related to this core g r w p of words in cllrster IV are sagengaw
(jedousylcompetitiveness) and niyefiyef (regret/angerat one’s self). Unlike the others, these
words focus on the self and what one could have done to prevent, or shoutd now do to
alleviate, the situation. Men often stated that they sorted these two emotions together
because they are both felt ”if someone goes fishing and catches a lot and you do not.”

w t i m word8 on Hrkrk 117

Niyefiyef will be felt if the reason for that discrepancy i s that one did not go fishing at all,
while sagengaw i s felt if one did go fishing but caught little. The two were also sorted
together as sequential emotions; one woman stated that if her mother died and had not yet
taught her how to weave, she would feel niyefiyef that she had not asked her mother to
teach her previously. She would then feel sagengaw and would try to learn to weave as well
as other women.
This latter pair of words and the more central words in the cluster were often linked
through such situations of differential fortune for the self and for ”other.” “If I feel
sagengaw to someone because they are smarter than me, 1’11 be both lingeringer and tip-
mochmoch with him and then gasechaula him.” Similarly, “If someone catches a lot of fish
you feel niyefiyef, and then you feel sagengaw to that person. You also become lingeringer,
which causes song. And then you gasechaula that person.“

cluster V the emotions of inability All of the words in this cluster are unified by their
common elicitation in situations where one is unsure about what to do or i s incapable of
doing what seems to be required by the situation. As a whole, this is the least unified of all
the five clusters. Several of the words in this group have strong connections with other
clusters; that is, many people sorted individual words in this cluster with words from other
clusters. Many of those same words were among the last to be sorted by people during the
task itself, as they were obviously less clearly members of any one particular group.
A close relationship is found between komayaya (incapable/indecisive), saumawal (con-
fused), and yeyewal (indecisive/doubtful). Most people said that they belong together
because all involve “not knowing something.” Some mentioned specific situations-most
particularly work situations-where one does not know what to do. The other words in the
cluster were variously sorted with this core group, or members of it, on the basis of situa-
tion. Some examples include the following:
If you are making thread, you can feel waires [worried/conflicted] if you don’t know if you can finish
it [in time]. You feel welingaw [suddenly bad] if you are not able to.
If you don’t like a certain kind of food, you feel both welingaw [suddenly bad] and niyabut
[disgusted] if you have to eat it.
If you are always doing things wrong, you feel saurnawal [confused] and filengaw [incapableluncom-
fortable].
When the teacher puts an assignment on the board at school, you look at it and feel all
three-nguch (sick and tiredlbored], saumawal, and waires.
Although the words in this cluster share the eliciting situation of a work or social obliga-
tion of which one i s unsure, or unable to fulfill properly, there are two subclusters within it,
labeled A and 6 in Figure I . These will be discussed below in reference to variation in
classification.

variation in classification
The five clusters to emerge from the sorting task reflect an “average” cognitive map of
this domain for the lfaluk in the sample. Although there was significant agreement between
individuals about the interrelations and similarities among these words, it is important to
examine in particular those emotion words that were sorted into widely different groups by
different individuals. Those links between words which are not reflected in Figure 1 will be
termed secondary classifications. They are important because they may reflect the use of
different dimensions of meaning of the words (for example, situation versus feeling tone, or
different aspects of a situation) by different individuals in their sorting. Secondary
classifications may also reflect ambivalence or ambiguity (White 1978) about the situations
in which a particular emotion is evoked. These classifications will be examined by cluster

118 amerlcan ethnologist

 and in terms of the meaning of the nonmodal pairings for the individuals who made them.
Cluster I,being the most distinct and different group, had only five links with other words
outside itself. These included ma (shamelembarrassment), lugumet (discomfort/guilt),
sagengaw (jealousylcompetitiveness), and fago (compassion/love/sadness). These links il-
lustrate that the good fortune that is the basis of cluster l may provoke ill feeling in others
(e.g., sagengaw). Responsibility for this ill feeling lies with the individuals who may have
“shown off” their good fortune. The person who was originally happy and excited then
feels ashamed (ma) or guilty (lugumet) for having flaunted his or her more favorable posi-
tion. Fago, which may be glossed as ”sad love,” is exclusively the love that i s felt for the
less fortunate. This is confirmed by the minimal linkage made between the happy emotions
of cluster I and fago, and contrasts with the strong similarities between the connotations of
love and happiness in English.s
The most important secondary classification not reflected in cluster II includes a number
of links made between itself and laloileng (insecurity). Although the latter word i s included
in cluster Ill with other emotion words having t o do with loss and connection, feelings of in-
security are elicited both in situations which are dangerous t o the self as well as in situa-
tions where the danger is to a significant other. The most important context in which danger
and potential loss occur i s during the illness of a relative.6 Laloileng i s also connected to
the other words in cluster V through their mutual elicitation in situations where one does
not know what will happen.
Cluster Ill i s linked to the emotion words in cluster IV through tang(frustration/grief). The
people who sorted rang and fago (compassion/Iove/sadness)together did so on the basis
that they are both felt in a situation where one feels for an unfortunate other but i s
somehow blocked through circumstance from acting on those compassionate feelings. As
one woman stated, “If someone i s sick or leaving the island, and I have nothing to give
them, I feel fago and tang.” On the other hand, the loss that causes fago may also lead to a
form of anger: “We fago someone who i s far away. We then feel nguch [sick and
tired/bored] at work and get song [justifiably angry] at people who talk to us while we are
thinking [of our fago for that person].”
Although nguch i s in cluster V with other words indicating uncertainty, it is also linked to
cluster Ill by mutual elicitation during separation from significant others. Nguch has even
stronger links t o cluster IV and in particular to song. Nguch is a much-used emotion word
that labels feelings in situations where one must accept that one’s individual goals are
thwarted. Although nguch and song are related to the blocking of goals, nguch is an ”un-
justified’’ or ambivalent anger. For this reason, it is not surprising that nguch was also
sorted with waires in cluster V. This i s in line with the use of nguch in daily life to describe
the frustration engendered by the obedience required to those of higher rank. While this
frustration cannot legitimately cause anger in one’s self, it is with some ambivalence and
difficulty that the label of nguch, rather than song, i s used in structurally similar (i.e.,
frustrating) situations.
The secondary classifications in cluster V include links between nguch, filengaw, we/-
ingaw, niyabut, and words outside the cluster. Filengaw (incapable/uncomfortable) has links
t o the emotions of ma (shamelembarrassment) and lugumet (discomfort/guilt) in cluster II
through their common elicitation in situations where one “stands out in the crowd” due t o
ineptitude. Welingaw (suddenly bad) is a more general term than most of the others. It i s a
“turning-of-the-stomach” feeling that may result from either disgust (niyabut) or the sudden
and simultaneous pull of conflicting definitions of the situation (waires). Welingaw may
also result from the same situations that cause song and therefore has links to cluster IV.
Niyabut (disgust) has the weakest links of any word in the domain with any particular other
word. Niyabut i s in fact only a ”quasi emotion,” as it refers exclusively t o the physical

emotlon words on lfaluk 119

revulsion experienced on seeing a spoiled or fetid object.’ The strongest link between
niyabut and others outside cluster V is that with gasechaula (hate). The latter emotion can,
like niyabut, also be felt toward a repulsive object.
Members of cluster V are loosely related to ma. In two closely related explanations, it
was said that “If there‘s gossip going around about me, 1’11 be ma, and then waires (wor-
ried/conflicted) about whether to go out or to hide at my house.” Another informant also
sorted these two words together on the basis of their cooccurrence when one tries to make
something for the first time: “1’11 be waires because 1’11 be ma that it will come out poorly,
and people will laugh.” Although niyabut and waires were never sorted together, a similar
situation t o those above was used to explain the sorting together of ma and niyabut: ”If our
clothes are old and people feel niyabut when they look at us, then we feel ma.”
What, then, do these clusters of emotion words reveal about the nature of emotions on
Ifaluk? Most significantly they show the importance of “situation” in the definition and
sorting of emotion words. Emotion is elicited in social interaction and in interaction with
environmental events. A person on lfaluk does not look inward to “discover” the emotional
state being experienced so much as he or she evaluates the existing situation. I do not mean
to imply that only situations are coded in these emotion words. In the foregoing analysis I
have noted instances where individuals explained their sorting on the basis of similarity of
internal feeling tone. While the evaluation of the environment must occur before a situa-
tion can truly ”elicit” a linguistically labeled emotion, in individual retrospect the strength
of a feeling may take precedence (in sorting) over situational similarity.
When a situation is evaluated on Ifaluk, it is in terms of i t s rewards for the self (cluster I);
dangers (cluster Ill; negative consequences for others and (through them) for the self
(cluster 111); frustration, through human error, of ego needs, including culturally constituted
needs (cluster IV); and, finally, comprehensibility or the potential for individual mastery of
the situation (cluster V). These broad evaluations of situations and the finer distinctions
made within each cluster have important reference to (1) cultural values that are being
either conformed t o or violated, and ( 2 ) the reactions of other individuals to the behavior of
the self within the situation. Individuals must appraise a situation as rewarding, punitive,
dangerous, frustrating, or overly complex within the constraints and with the aid of cultural
values, as well as in relation to significant others, in order to label themselves with a
specific lfalukian emotion word.

Multidimensional scaling i s less responsive than hierarchical clustering to cultural and

linguistic nuances of meaning, but it can give a more holistic picture of the dimensions of
meaning and of the relations between words on those dimensions. The same similarity
matrix of the sorted emotion words used in the clustering analysis was submitted to a
multidimensional scaling procedure. Figure 2 presents the two-dimensional results.’ A
three-dimensional model resulted in only slight reduction of stress and will not be con-
sidered here (stress = ,0882 for the two-dimensional model and .0507 for the three-
dimensional model).
The clearest and most striking dimension along which the emotion words fall is what I
have called pleasant versus unpleasant effects of the emotion-eliciting situation. This
dimension sharply divides cluster I from all the others and is similar to the “evaluation”
dimension that Osgood (Osgood, May, and Miron 1975) found in his cross-cultural studies
of meaning. Of interest here i s what the lfaluk evaluate when they sort their emotion
words. It is necessary to return to the cultural data that are the source of a multidimen-
sional scaling in order to name and explain the dimensions. While it appears to be univer-
weakness. It may be the result of moral rectitude (strength) or wrongdoing (weakness), or it
may reflect an individual's possession of higher social rank, more resources, or greater
cognitive control or understanding of the situation (strength)-or the lack of the same
(weakness). Thus, this second dimension refers to ego's encounter with the environment,
and the appropriate emotional response is one that takes account of one's position in rela-
tion to it.
The emotions of danger (cluster 11) and the emotions of inability (cluster V) are both found
at the "weak position" end of the dimension and share an uncertainty about performance
or about the consequences of the situation that elicits them. The emotions that arise out of
a "strong position" include those emotions that entail certainty about oneself in relation to
the environment. The emotions of connection and loss (cluster Ill) and the emotions of
human error (cluster IV), found at the "strong position" end of the dimension, are generally
emotions that are felt toward an individual who is "weaker" than oneself, either through ill-
ness or other misfortune (for example, fago [compassionllovelsadness]), or through par-
ticipation in a morally reprehensible act (for example, song [justified anger] ).
This dimension does not refer to the strength or weakness of the feeling tone which ac-
companies each emotion. Two of the emotions that the lfaluk say produce the most disrup-
tive physiological symptoms-rus (paniclfrightlsurprisel and niyabut (disgust)-are found
at the "weak" end of this dimension. Similarly, cluster I words (the emotions of good for-
tune) include reference to excited hyperactivity and have "strong" behavioral expressions
such as shouting or running, but they are located near the midpoint of the dimension. in ad-
dition, cluster I emotions may cause one to misbehave and lead to "jealousy" in others.
This indicates that the second dimension refers to interpersonal relationships of strength
and weakness more than to an atomistically defined strength, such as might emerge from
the possession of material goods.'
Although this dimension parallels that of "potency"(0sgood et al. 1975),it takes i t s par-
ticular meaning on lfaluk in reference to the institutions and values of rank. In a political
and economic system that is based on individual achievement and that encourages the dif-
ferential distribution of resources, "potency" may carry stronger connotations of superiori-
ty and unidirectional control. On Ifaluk, by contrast, rank carries the responsibility of
benevolence and nurturance toward those "lower" or more "needy" than oneself. The or-
thogonality of the dimensions based on evaluation and potency (Figure 2) indicates that the
lfaluk do, as their values insist, differentiate rather sharply between strength and self-
aggrandizement. This is not meant to imply that there are no ego rewards to the altruism
that ideally accompanies strength on Ifaluk. Rather, it i s exactly the cultural molding and
social desirability of an emotion like fago (compassion/love/sadness)that gives this dimen-
sion the meaning attributed to it here.
In definitions of emotion words, in conversational use, and in explaining their sorting
decisions, the lfaluk primarily define the emotions by the situations during which they are
felt, and this i s reflected in the foregoing analysis. The two most important dimensions of
emotion-eliciting situations are the rewards that they hold for ego (horizontal dimension)
and the strength of ego relative to the people and objects in the situation (vertical dimen-
sion). Figure 3 illustrates this with the situational elicitors (as revealed in folk definitions
and in the uses of the terms in everyday conversations) for some of the emotions located at
outlying points in the dimensional space defined by the scaling procedure.
The emotions at the top of Figure 3 are generally felt by individuals of higher rank
toward those of lower rank. An elder is more often song (justifiably angry) at a younger per-
son than at a peer or at a higher-ranked individual. The chiefs are often said to be song at
these who have broken rules or taboos. Fago i s more frequently directed from parents to
their children than in the other direction, and the ideal chief i s one who feels fago for his

122 rmerican ethnologlst

 STRONG POSITION

A 50NG (justified anqor) A FAG (compassion /bve/rodness)

-the other b r d S 0 tObW Or -the othw is ill or hunqry
viololn o norm -the othw must k v a Itm blond
-M~olherisob.d~ntandcolm

’
s
W

E
W
tW
AGASECHAULA(hate)
-thoother c a u m harm to epo
(theft, poasip,etc.)
-tha othr’shows off’,is
A KER (hoppiness/e~citemcnt)
-ego ptsfoodorgoob

AbACNU (likinq)
-theothr is pkasont,
f

E
friendly
k- aqqmsive -090 dwivss pleawra
fromfoodorgoods

3
A KOYAYAYA (inc~ability/indrcision)
- q o 4 a n not know how to
a
complete lash
AMETAGi Ihor/onaiety)
-thaothris song bgo has mistmhavedl
-IkeOn)eIiso qhwt
-tne dherisuntomilior to 090

WEAK POSITION

Figure 3. Situational elicitors of emotions a t outlying points in two dimensions

subjects. Note that gasechaula (hate) may occur in the same situations that elicit song, but
the former i s more often used between peers, and even here only rarely. This i s reflected in
the place of gasechaula at a midpoint between the strong and the weak position ends of the
vertical dimension.
In a dyadic interaction, the song of one person (usually of higher rank) produces the emo-
tion of rnetagu (fear/anxiety) in the other (usually of lower rank). Metagu and kornayaya (in-
capability/indecisiveness) are felt primarily in situations in which ego has done something
morally wrong or performed incapably. To the right in Figure 3 are two emotions that are
both spatially and situationally orthogonal t o the system of rank that informs the vertical
dimension. Ker (happinesslexcitement) occurs in situations where the other i s neither song
nor rnetagu, but this fact makes ker a somewhat dangerous, “uncontrolled” emotion. Like
gasechaula, gachu (like) is most often used to describe one’s feelings toward peers. Ker and
gasechaula hold extreme positions at the ends of the horizontal dimension, in contrast to
their shared orthogonality to the vertical dimension. As can be seen in Figure 3, ego i s
punished by the situations that elicit gasechaula and i s rewarded by those that elicit ker.

discussion
The lfalukian cognitive organization of the domain of 31 emotion words is based on the
eliciting situation. The five major groupings of emotion words in the hierarchical clustering
analysis were named according t o five basic situation types: good fortune, danger, loss and
connection with others, human error, and overly complex and misunderstood events. The
multidimensional scaling procedure pointed out two major dimensions of emotion words
that were also interpreted in terms of eliciting situation. Those dimensions include the
pleasant or unpleasant nature of the consequences of the situation for ego and the strength
or weakness of ego in relation to the other. Stressed throughout i s the meaningfulness of
the groupings and dimensions for lfalukian individuals. Equally emphasized i s the articula-
tion of the dimensions of meaning with dimensions that are present in lfalukian values and
social institutions. It is these cultural factors that structure and give meaning to the situa-

emotion words on lfaluk 123

tions which are associated with each emotion and which in fact create the emotional mean-
ing of the situation.
The similarity between the dimensions found here and those found by several other
researchers across various types of semantic domains is clear. Osgood (Osgood et al. 1975)
has described the dimensions of "evaluation, potency, and activity" as universal aspects of
meaning. White (19801, in a comparison of the scaling of Melanesian and Indian personality
descriptives, finds "solidaritylconfl ict" and "dominancelsubmission" to be the most impor-
tant dimensions. He relates them to Osgood's evaluation and potency dimensions, respec-
tively. There have been numerous analyses of English emotion words which find some of
these same dimensions."
The evidence from other domains and other cultures suggests a universal tendency to
perceive "the meaning of things" with broadly similar criteria. It i s crucial, however, to ex-
amine what in particular i s the focus of perception. This focus will vary in similar domains
across cultures and across semantic domains within particular cultures. Most importantly,
this variation may be very informative about both culturally framed perceptions (for exam-
ple, lfalukian views of emotion versus Samoan views of emotion) and the domain in ques-
tion (for example, emotion words or personality trait terms). These two types of variation in
the focus of perception are briefly examined in conclusion.
In terms of variation across cultures within the same domain, the dimensional analyses
of English emotion words mentioned above are very instructive. A comparison of these
analyses reveals that the two most common dimensions found are the pleasantnesslun-
pleasantness of the feeling state and the level of activation. The first dimension i s called
"hedonic tone" (comfortldiscomfort) by Davitz (1969) and "pleasantlunpleasant" by others
(e.g., Schlosberg 1954; Block 1957; Frijda and Philipszoon 1963). The second most common
dimension i s "level of activation," which refers to physiological and/or behavioral activity
(Schlosberg 1954; Block 1957; Abelson and Sermat 1962;Davitz 1969 Bush 1973). Although
these and other investigators have posited some secondary dimensions of emotion that
relate to the position of ego in relation to the environment," the focus of perception in
emotional situations i s presumed to be primarily internal.
The focus of these latter studies on the psychological, physiohical, and, more general-
ly, internal dimensions of emotion contrasts with the lfalukian focus on the external dimen-
sions of emotion, that is, on the eliciting situation. Cultural differences m the meaning d
the dimensions of evaluation and potency are related, in this case, to ethnopsychological
beliefs about the nature of emotion and self. While Americans define e m o t i o n s primarily as
internal feeling states (see also Cardiner e t al. 1937; Solornon 1976,19711). the lfaluk see the
emotions as evoked in, and inseparable from, social activity (Lutz 1seO).
Another aspect of focus is illustrated by the organization d the domain of emotion
words in Samoan (Gerber 1975). The dimension of evaluation is central to Samoan emotion
words as well as to American English and lfalukian terms. This dimension is focused on
pleasant and unpleasant feeling states in the English case, and on the pleasant or unpleas-
ant consequences of the situation on Ifaluk. An analysis d sorted SarncMIl emotion words
reveals that the focus of evaluation there, however, i s on the social aood rather than cm
pleasant feeling or consequences.The positively evaluated emotions are the ethically cw-
rect emotions. These "socially virtuous" emotions are, moreuver, strongly associated with
submission, and they are considered murc unpleasant than pleasant to expcrieme.
The extent of overlap between socially valued emotions and pkasant or - m i c a b
rewarding emotioAs is thu5 culturally v w k b k , as is the extent te which m e or the & is
seen as central to the meaning of emotion words. The "pursuit of happiness" may be a
cukwal value and i s portrayed as such, far exam+, k American public w. In
lfalukim belief, the ~ o u of g ethically c m c t or good cmotierrs i s not the same as the
 group of emotions that occur in situations bringing personal reward. Involvement in a
rewarding situation is not always, or perhaps even usually, culturally sanctioned. Variation
in the focus of evaluation in Samoan, Ifalukian, and American emotion word meanings can
shed important light on the effects of the cultural values and theories that inform the
perception of emotion.
Variation across semantic domains in the focus of meaning is a second important source
of information. One would then ask how the meaning of emotion words differs from, for ex-
ample, that of personality trait terms. Several investigators have noted a close relationship both
between states and traits themselves (Izard 1977) and between emotion terms and per-
sonality trait terms (Plutchik 1980). Both sets of words can be used to describe the
characteristics of individuals, with emotion words focusing on more temporary than perma-
nent characteristics. Both types of words are also interpersonal judgments which, as White
(1980:776)points out, rely on common “conceptual themes in folk interpretations of social
behavior generally.”
The difference between the dimensions found by White to inform personality descrip-
tives (“solidarity/conflict” and “dominance/submission”) and those found here for emotion
words are instructive about the nature of these two domains. White (1980:767)sees these
dimensions as based on ”inferential knowledge,” which is much more complex than the
“simple connotations of ‘goodness’ and ’strength‘.” The dimension of solidarity/conflict
emerges from the possibility for matching or mismatching between the goals of two actors,
while the dimension of dominance/submission arises from the fact that individuals in-
fluence each other. Personality terms are most frequently used to predict or explain the
behavior and presumed motivations of the other (White 1980:767).
Although a predictable other is one who can be counted on to either support or block ego’s
goals, the focus is on the other. Psychological studies among American populations have
shown, in fact, that people observing the behavior of others tend to see the actor as ”figural
against the ground of the situation” [Jonesand Nisbett 1971:lS).In describing the causes of
their own behavior, however, actors speak in terms of more immediate situational
demands. Trait descriptors would therefore be used most frequently to depict others rather
than self.
Emotion words, by contrast, simultaneously code changing situations and ego’s stance in
relation to those situations. Although it has been seen that the primary referent of emotion
words may be either physiological or external events, these words are more readily used to
describe the state of the self (either alone or in relation to the environment) than are per-
sonality trait terms.
Finally, we may ask about the relation between emotion words and emotion itself. Emo-
tion words are used for giving accounts of experience. The extent to which culture provides
conceptual and ethnotheoretical tools for constructing emotion experience itself, as well
as accounts, has not as yet been fully realized. In part this is due to the influence of Western
views of emotion as an essentially private, unlearned, and physiological process. There is
evidence accumulating in other domains that the personal and cultural theories that peo-
ple use in constructing accounts of experience have powerful, direct, and generative ef-
fects on such things as problem solving (Centner 198l),the perception of and action within
a marital relationship (Quinn 198l),and the production of metaphor (Lakoff and Johnson
1930).
Divergence between what people say and what they presumably know (and feel) has
been presented as the central flaw in research that uses verbal reports (Nisbett and Wilson
1977) a d in many studies in cognitive anthropology (Laboratory of Human Cognition
1978:63-66).Rather than dismissing verbal reperts as inaccurate, however, it would seem
more useful to investigate “the transformations that lead to introspective accounts”
(Mandler 1980242) and, in particular, cultural conventions for communicating personal ex-
perience t o others o r for concealing it (LeVine 1981:5). Explicitly examining the various con-
texts in which emotion words are used, including structured a n d unstructured interviews
with an ethnographer, and use in conversations with a wide variety of others can move us in
the direction of understanding the relation between emotion words and emotion ex-
perience.
The distinctive nature of emotion as a phenomenon lies in the f a c t that it appears to the
individual to originate both in the self and in the world. Cultural variation exists in the ex-
tent to which either the former o r t h e latter is emphasized. In either case, emotion words do
not simply serve to bring the private into the social realm. The cultural values a n d theories
that inform the meanings of emotion words give those words an important role in aiding in-
dividuals in the interpretation and understanding of their situated selves.

notes
Acknowledgments. The data on which this paper is based were collected in 1977 and 1978 during 12
months of fieldwork on the atoll of lfaluk in the Western Caroline Islands. The research was supported
by a grant from the National Institute for Mental Health. I am grateful to John Whiting, Beatrice
Whiting, Robert LeVine, Martin Etter, Geoffrey White, the late Michelle Rosaldo, and anonymous A €
reviewers for their constructive comments on drafts of this paper. An earlier version of this paper
received the 1980 Stirling Award in Culture and Personality Studies.
' Further ethnographic description of lfaluk is available in Burrows and Spiro (1953). Burrows
(19631, and Lutz (1980).
* These words were cross-checked with several other informants by asking them to free associate to
a cue word from the list. This process added few words to the list and consisted miinly of terms
already acquired by the other methods mentioned here. It i s certain, however, that there are rarely
used emotion words that were not discovered here.
The sorting task was performed by six females and seven males, of whom all but one were
literate. The sorting of that one individual did not differ significantly from the others and is included in
the analysis. Literacy of at least a minimal sort is nearly universal among islanders under the age of 40.
The task seemed to be easy as well as enjoyable, and most people completed it within a half hour.
' The program used was Cluster Analysis PIM, designed by John Hartigan and described in Dixon
and Brown (1977). An "index of structure" for the domain was computed following the method
presented by Von Glascoe (1979:308).
The two pairings of fago (compassion/love/sadness)and ker (happinesslexcitement)were made by
men with some knowledge of English. On questioning, these men translated fago as "love." The
possibility that some sortings were made on the basis of a translation into English is problematic, but it
may point to important structuring effects of language.
Metagu (fearlanxiety), rus (panic/fright/surprise),and laloileng (insecurity) were sorted together in
several cases. One young man said that he feels all three emotions in relation to the possibility that his
father may fall out of a coconut tree. One person sorted them together because they can be elicited se-
quentially in a situation: laloileng is felt when one does not know what will happen in the future, which
then causes rus; one is then metagu of the thing which has caused one's rus.
' Niyabut (disgust) was kept in the sample despite the statements by several informants that it is "a
bit different" from the other words. This was done in order to allow comparison with "disgust," which
is considered by many to be a universal emotion (Tomkins 1963; Ekman 1974).
The scaling program used was KYST, designed by Kruskal, Young, and Seery (1973).
Note that gachu (like) and chegas (romantic pride), both at the "weak position" side of the second
dimension, refer more than the others in that cluster to interpersonal satisfactions. To "like" or be
"proud" of another may involve an element of weakness vis-his the other.
"The data on which these dimensional analyses have been based vary from coded interview
material and judgments of facial expression through introspection and post hoc theorizing. In any case, they
reflect an important aspect of American (both scientific and lay) cognitive organization of emotion words.
'' One such posited social dimension that relates to the dimension of strength/weakness found in
the present study has been variously but similarly labeled by several individuals "social submission/
condescension" (Frijda and Philipszoon 1963), "feel superior/feel inferior" (Krech and Crutchfield
1965). "competence" (Davitz 19691, and "dominance/submission" (Russell and Mehrabian 1977).

126 amerlcsn ethnologist

 references cited
Abelson, Robert P., and Vello Sermat
1962 Multidimensional Scaling of Facial Expressions. Journal of Experimental Psychology 63:546-554.
Block, Jack
1957 Studies in the Phenomenology of Emotions. Journal of Abnormal and Social Psychology 54:358-
363.
Burrows. Edwin C.
1963 Flower in My Ear. Seattle: University of Washington Press.
Burrows, Edwin G., and Melford Spiro
1953 An Atoll Culture: Ethnography of lfaluk in the Central Carolines. New Haven: HRAF.
Bush, Lynn E.
1973 Individual Differences MDS of Adjectives Denoting Feelings. Journal of Personality and Social
Psychology 25:50-57.
Davitz, Joel R.
1969 The Language of Emotion. New York: Academic Press.
Dixon, W. J., and M. 8. Brown, eds.
1977 Biomedical Computer Programs P-Series. Berkeley: .University of California Press.
Ekman, Paul
1974 Universal Facial Expressions of Emotion. In Culture and Personality: Contemporary Readings.
Robert LeVine, ed. pp. 8-15.Chicago: Aldine.
Frijda, Nico H., and Els Philipszoon
1963 Dimensions of Recognition of Expression. Journal of Abnormal and Social Psychology 66:45-51.
Gardiner, Harry, Ruth Metcalf, and John Beebe-Center
1937 Feeling and Emotion: A History of Theories. New York: American Book Company.
Centner, Dedre
1981 Generative Analogies as Mental Models. In Proceedings of the Third Annual Conference of the
Cognitive Science Society. pp. 97-100.Berkeley.
Cerber, Eleanor
1975 The Cultural Patterning of Emotions in Samoa. Ph.D. dissertation. Anthropology Department,
University of California, San Diego.
Izard, Carroll E.
1977 Human Emotions. New York: Plenum Press.
Jones, Edward E., and Richard E. Nisbett
1971 The Actor and the Observer: Divergent Perceptions of the Causes of Behavior. Morristown, NJ:
General Learning Press.
Kagan, Jerome
1978 On Emotion and Its Development: A Working Paper. In The Development of Affect. Michael
Lewis and Leonard Rosenblum, eds. pp. 11-41.New York: Plenum Press.
Krech, David, and Richard 5. Crutchfield
1965 Elements of Psychology. New York: Knopf.
Kruskal, Joseph, Forrest Young, and Judith Seery
1973 How to Use KYST: A Very Flexible Program to Do Multidimensional Scaling and Unfolding. Mur-
ray Hill, NJ: Bell Laboratory.
Laboratory of Comparative Human Cognition
1978 Cognition as a Residual Category in Anthropology. Annual Review of Anthropology 7:51-69.
Lakoff, George, and Mark Johnson
1980 Metaphors We Live By. Chicago: University of Chicago Press.
LeVine, Robert
1980 Anthropology and Child Development. In Anthropological Perspectives on Child Development.
Charles M. Super and Sara Harkness, eds. pp. 71-86.San Francisco: Jossey Bass.
1981 The Self in Culture: PersonCentered Ethnography and Psychoanalytic Anthropology. Ms. Files
of the author.
Lutz, Catherine
1980 Emotion Words and Emotional Development on lfaluk Atoll. Ph.D. dissertation. Anthropology
Department, Harvard University.
1981a Talking About "Our Insides": lfalukian Conceptions of the Self. Paper presented at the Annual
Meeting of the Association for Social Anthropology in Oceania, February 1981,San Diego.
1981b Situation Based Emotion Frames and the Cultural Construction of Emotions. In Proceedings
of the Third Annual Conference of the Cognitive Science Society. pp. 84-89.Berkeley.
Mandler, George
1980 The Generation of Emotion: A Psychological Theory. In Emotion: Theory, Research, and Experi-
ence. Robert Plutchik and Henry Kellerman, eds. pp, 219-243.New York: Academic Press.

emotion words on lfaluk 127

Myers, Fred
1979 Emotions and the Self: A Theory of Personhood and Political Order among Pintupi Aborigines.
Ethos 7:343-370.
Nisbett, Richard E., and Timothy D. Wilson
1977 Telling More Than We Can Know: Verbal Reports on Mental Processes. Psychological Review 8 4
231-259.
Osgood, Charles, William H. May, and Murray 5. Miron
1975 CrossCultural Universals of Affective Meaning. Urbana: University of Illinois Press.
Plutchik, Robert
1980 A General Psychoevolutionary Theory of Emotion. In Emotion: Theory, Research, and Experience.
Robert Plutchik and Henry Kellerman, eds. pp. 3-33.New York: Academic Press.
Quinn, Naomi
1981 "Commitment" in American Marriage: Analysis of a Key Word. Ms. Files of the author.
Russell, James A., and Albert Mehrabian
1977 Evidence for a ThreeFactor Theory of Emotions. Journal of Research in Personality 11:273-294.
Schlosberg, Harold
1954 Three Dimensions of Emotion. Psychological Review 61:81-88.
Sohn. Homin, and Anthony Tawerilmang
1976 Woleaian-English Dictionary. Honolulu: University Press of Hawaii.
Solomon, Robert C.
1976 The Passions. New York: Doubleday-Anchor.
1978 Emotions and Anthropology: The Logic of Emotional World Views. Inquiry 21:181-199.
Tomkins, Silvan S.
1963 Affect, Imagery and Consciousness. Vol. 2:The Negative Affects. New York: Springer.
Von Glascoe, Christine
1979 Evidence for Multiple Cognitive Realities in Yucatec Game Cognition. In Ethnolinguistics: Boas,
Sapir and Whorf Revisited. Madeleine Mathiot, ed. pp. 297-312.New York: Mouton Publishers.
White, Geoffrey
1978 Ambiguity and Ambivalence in A'ara Personality Descriptors. American Ethnologist 5334-360.
1980 Conceptual Universals in Interpersonal Language. American Anthropologist 82:759-781
1981 "Person" and "Emotion" in Aara Ethnopsychology. Paper presented at the Annual Meeting of the
Association for Social Anthropology in Oceania, February 1981,San Diego.

Submitted 20 February 1981

Accepted 15 April 1981
Revised version received 26 October 1981

128 amerlcan athnologlst