LANGUAGE, CONCEPTS, AND EMOTIONS


The Role of Language in the Construction of Emotions

Michael Bamberg

0. Introduction

The relationship between language and emotions can be viewed from two angles. First, language, in a broad sense, can be viewed as being done [performed] "emotive". Taking this angle, it is commonly assumed that people, at least on occasions, "have" emotions, and that "being emotional" gains its own agency, impacting in a variety of ways on the communicative situation. 2 This can take place extralinguistically (e.g. by facial expressions, body postures, proximity, and the like), in terms of suprasegmentational and prosodic features, and in terms of linguistic (lexical and syntactic) forms. A recent collection of articles in a special issue of the Journal of Pragmatics (Caffi & Janney 1994; see also Fiehler 1990, Bamberg & Reilly in press) testifies to this research orientation. Although research along this line of reasoning focuses primarily on the "expression" of emotions, i.e. the behavioral act of expressing affect in communication, it nevertheless relies heavily on (often culturally privileged - see Besnier 1994) notions of what emotions are and how they function in private and public settings. In this view, language and emotion are two concurrent, parallel systems in use, and their relationship exists in that one system (emotions) impacts on the performance of the other (language). Both of them share their functionality in the communicative process between people.

    The other tack on the relationship between language and emotion inverts the directionality of the view just discussed. It starts from the assumption that language in a way refers to, and therefore "reflects" objects in the world, among them the emotions: Languages have emotion terms, and people across the world engage in talk about the emotions - though not necessarily to the same degree and with the same obsession and reflexivity as in the so-called Western world. In this view it remains unspecified whether emotions are 'real' objects in the world such as behaviors or whether they are 'internal' psychological states or processes (resembling other psychological processes such as thoughts or intentions). This view then takes a different tack to the language-emotion relationship. Language is a means of making sense of emotions, and as such can be used as a starting point to explore the world of emotions in different languages as well as in different "language games".

    However, taking this orientation as the starting point, one is immediately challenged to consider the role of language in much more detail. If language is conceived of as merely representing (in the sense of 'mirroring') the world of emotions and/or people's conceptualizations and understandings of the emotions, language offers an immediate access. Language, in this view, is "transparent". If language, however, is conceived of in one or another way as contributing to how emotions are understood, or even, to what emotions "are", the relationship is not direct, but mediated.

    It is this second orientation that I will take as a starting point for this paper. First, I will, in an admittedly rather eclectic fashion, discuss three approaches that revolve around language issues as a starting point to explore emotions (section 1). I selected these three different approaches for two reasons: First, they start from quite different assumptions of what language is, how it functions, and in addition, with regard to its transparency. Examining the assumptions that lie behind the individual approaches will help reveal some of the background that led to my own "linguistic-constructionist" approach 3 . Second, although I am somewhat critical of all three theoretical frameworks discussed, they have been (and still are) the most appealing to me, in as far as they were most influential in my own thinking after my interest in the relationship between emotions and language had been spurred by two of my mentors, George Lakoff and Dick Lazarus, during my graduate training in Berkeley. After having taken critical account of the three approaches, particularly with their underlying assumptions regarding the role of language and the approach to development invoked, I will turn in section 2 to a summary of some of my own findings. These originated from a project that was funded by The Spencer Foundation, having led me to see the need to continue this line of research with a stronger emphasis on cross- cultural comparisons. In the last section of this paper (section 3) I will turn to some more methodological and theoretical considerations with regard to the relationship of language and emotions, opening up the central issue of the role of language in the appropriation of the emotions and in their development.

1. Language as a tool to explore emotions

1.1. Anna Wierzbicka's "universal semantics"

In numerous articles, chapters and books Wierzbicka has explicated her theoretical stance on how to analyze emotions. Emotions to her are a semantic domain (1995a: 235), to be investigated in a semantic metalanguage, i.e. in terms of indefinables or primitives (semantic universals) that are shared by all human languages. These universals are of a conceptual nature and comprise elements such as feel, want, say, think, know, good, bad, and so on (1992: 236; 1994: 140; 1995a: 236). It is Wierzbicka's declared aim "to explore human emotions (or any other conceptual domain) from a universal, language-independent perspective" (1995a: 236).

    In her comparative study of language-dependent conceptualizations, Wierzbicka is able to document that "every language imposes its own classification upon human emotional experiences, and English words such as anger or sadness are cultural artifacts of the English language, not culture-free analytical tools" (1992: 456; 1995a: 236). Her analyses are good (and clear) examples for exactly this point, and her main argument is forcefully directed against most psychological theorizing within the James-Lange-tradition that starts from the assumption that emotions are bodily experienced feeling states, each categorically distinct, and built up in a clearly ordered sequence of events (see for recent critiques of this kind of theorizing from within psychology, though from quite different directions, Campos, Mumme, Kermoian & Campos 1994, Ellsworth 1994, Sarbin 1995).

        While the suggested set of semantic primitives that is assumed to exist in every human language started out with only fourteen, it is currently estimated (Wierzbicka 1995b, Goddard in press) to have increased to about 35-60 elements. In the following two examples of the semantic explications of the abstract concept "GUILT" and the concrete concept "SKY" (from Wierzbicka 1995b: 293), all the terms mentioned in combination are supposed to be universals and as such parts of what has been termed the "Natural Semantic Metalanguage" (NSM):

    X felt guilty =

    X felt something

    sometimes a person thinks something like this:

        I did something

        because of this, something bad happened

    because of this, this person feels something bad

    X felt like this

    sky

    it is above everything

    it is above all places

While the above explications resemble previous explications of situated and culturally shared meanings developed by Geertz (1973), Labov and Fanshel (1977), Much (1992) or Shweder (1991), it needs to be stressed that the explications within the NSM-framework are argued to operate from a non-contextual, culture free starting point. The linguistic ideology evoked not only is that this objective starting point exists [in the form of the NSM], but also that the human mind is innately equipped with it.

1.2. Rom Harré's "emotionology"

Harré's suggestion that researchers study "the way people use their emotion vocabulary, in commenting upon, describing, and reprimanding people for emotional displays and feelings" (Harré & Gillett 1994: 148), is in many ways similar to Wierzbicka's approach. In aiming to pull out of the uses of the emotion vocabulary (of a given culture at a given time) the underlying "theory of emotion", Harré and Gillett follow Stearns and Stearns' (1986, 1988) theory of "emotionology". In contrast to the universal orientation of Wierzbicka, an emotionology is a very local theory (and taxonomy), which is said to consist of four general features. These features need close attention if an emotion is to be identified and labeled correctly: (1) a felt bodily disturbance, (2) a characteristic display, (3) the expression of a judgment, and (4) a particular illocutionary force.

    In taking the work of Catherine Lutz (1988) on the Ifaluk emotionology as an example, Harré and Gillett document that emotion words function to fashion - so to speak - emotional acts. This is not meant to imply that emotionologies are the same as emotion display systems; but describing a person as angry, or in terms of the Ifaluk as doing metagu ('behaving inappropriately'), passes a judgment on the person talked about, and implicates this person with regard to the performance of a particular illocutionary act. Thus, emotion words do the job of orienting toward a particular "positioning" (usually of those who are characterized by these terms) within social encounters.

    Thus, while at first sight Harré's suggestion to study emotion terms as part of a (decontextualized) lexicon sounds very much like Wierzbicka's study of word reference, his proposal to study the emotion vocabulary of a certain group of people promises a whole lot more. If references to emotions - in the realm of texts - position people with regard to one another (morally, aesthetically, and prudentially), they function as indexes to how emotions in discourse situations are displayed, that is, what emotions mean as discursive acts. Thus, employing emotionologies for the study of emotions as discursive acts orients us toward the study of language use in discourse situations. Stearns (1994, 1995) in coining the term "emotional styles" - I assume somewhat under the influence of Geertz's notion of "cultural styles" - draws attention to the study of culture, though not by venturing into general hypotheses about universal functions and characteristics of emotions. Rather, culture means "gaining a comparative sense - how the emotion in question is different from one case to the next - even if the end result is primarily to understand one particular cultural expression..." (Stearns 1995: 45).

        An interesting point for venturing into the acquisition or the learning of emotionologies lies in Harrés affinity and flirtation with anthropological approaches to emotions which traditionally place emphasis on the cognitive and/or conceptual aspects of emotions (cf. Lutz & White, 1986; Lutz, 1988). In an apparent parallel with Wierzbicka, Lutz claims that "emotion can be viewed as a cultural and interpersonal process of naming, justifying, and persuading by people in relationship to each other" (Lutz 1988: 5). However, in contra- distinction to the former, the anthropology of emotions does not seem to claim psychological reality for the kind of conceptual analysis that is employed in the process of explicating the experiential, expressive, and regulative aspects of emotion displays. However, this leaves the question open, as Oatley formulated in his review of Lutz's works, "of what emotions might be constructed from" (Oatley 1991: 77), and how much this construction process can (and needs to) be described in conceptual/cognitive terms, and what role language and other symbolic systems might play in this construction process.

1.3. Nancy Stein's theory of "goal-action-outcome knowledge"

In a number of publications spanning over at least the last two decades, Nancy Stein and her associates have been investigating the cognitive capacity to simulate the plans and goals relevant to the understanding of (human) actions as part of the study of personal and social behavior (e.g. Stein 1979, 1988, Stein & Albro in press, Stein & Levine 1990, Stein & Trabasso 1992). In this approach, knowledge of goals and plans is assumed to be the basic prerequisite to make sense of others, and it figures in the same capacity in explaining and reasoning about one's own actions, i.e. in the process of making sense of one's own self. Stein's original research in children's understanding of human intentionality in their story constructs has recently moved focus more strongly on the appraisal processes relevant to assessing the specific goals, values, and moral principles involved in the understanding of characters' actions. And since knowledge about valued goals and their outcomes is taken to be heavily influential on how characters 'feel' and how they (con-sequentially) react, a model has emerged that is argued to capture the "meaning of emotions", i.e. it can spell out the appraisal processes that link the occurrence of a situation to an emotion (and potential subsequent reactions), and delineate the nature and boundaries between emotions (Stein & Trabasso 1992, Stein, Trabasso & Liwag 1993).

    Making use of a particular narrative interview technique, labelled the "on-line interview", individuals are continually monitored in their reporting of an emotional experience with regard to the status of their goals and valued preferences. Since it is assumed that goals and action plans undergo changes within actual emotional experiences, this technique is supposed to tap these changes and thus facilitate to get 'beyond' the report of the experience 4 . The questions asked to gain these insights into the components of appraisal processes are: (1) what happened?, (2) what can be done about it? [in first person scenarios this is further differentiated into (a) what can I do about it?, and (b) what did I do about it?], and (3) what are the likely results of carrying out a plan of action? (Stein, Trabasso & Liwag 1994, Stein, Liwag & Wade in press).

    Thus, while Wierzbicka views emotions as a semantic [= conceptual] domain which governs the patterns of discourse, and Harré takes emotions to be part of the domain of statements (= actions and interactions), Stein's approach to emotions - to a degree at least - seems to combine aspects of both: Emotions are schematically organized, i.e. part of a representational system, and these schemata are 'put to work' in responses to emotional events in the form of being angry or doing metagu. However, they first of all are cognitions, constituting the motivational force for individuals to (re)act in a certain manner. It should be noted that the "goal-action-outcome" theory claims to be able to account for cultural (and individual) variations by decomposing the general intentional states into distinct components which can be filled in and arranged in culturally variable ways.

1.4. Language, culture and cognition: How are emotions learned?

When, as in Stein's approach, emotions are approached as a representational system of some specific goal-plan-outcome knowledge, then the acquisition of categorial distinctions between the (basic as well as culture-specific) emotions consists in its most basic form of knowledge of intentional action and of goal plans. According to Stein and her associates, these two knowledge types are acquired relatively early, at around three years of age. At this point, children can successfully differentiate the components that lead to (English) anger, sadness, fear, and happiness (cf. Stein, Liwag & Wade in press), 5 and efforts have concentrated on illustrating how children make use of this knowledge in later years in other problem solving tasks (see for instance: Trabasso, Stein, Rodkin, Munger & Baughn 1992). Using narratives of real life (emotional) situations and subjecting them to on-line questions for on-line reasoning, Stein and her colleagues rely on language in its ideational, representational function, as a relatively transparent window to what the narrator means when talking about emotions. The content of the topic is taken as what is of basic concern, and whether the narrator wants to be understood as blaming another person or saving face, i.e. the directive force of language use [= the interpersonal function], is not considered of immediate relevance. Similarly, the directive force of culture in the form of rules and norms for conduct only becomes relevant as a 'secondary' force, i.e. by way of individually represented cognitive systems, containing information about individual experiences and culturally shared knowledge.

    Within the framework chosen by Wierzbicka, language should obviously play a much more important role, since her system of universals is labelled 'semantic primitives', which, however, in their very nature are considered conceptual [= cognitive]. Although, to the best of my knowledge, Wierzbicka did not investigate how these semantic/cognitive universals are acquired, there are two options: Option number one consists of a nativist/maturational solution. Accordingly, we bring the set of culture-independent universals to the process of cognitive development, and out of these, the child has to narrow down the options that are chosen by the specific culture and "encoded" by the specific language in specific linguistic (lexical) forms. Option number two would argue the other way around: The set of universals is 'learned' by way of abstracting them from the language specific forms (and their local meanings) which are learnt first. According to this version a culture-independent language could be the product of a reflective learning process that operates on top of a first language, that is after having developed a more detached position from one's own culture and language use. Although I would be leaning toward this version, Wierzbicka seems to strongly favor the assumption that the universals are innate (Wierzbicka 1995b), particularly, since version two would downgrade her set of universals to linguistic constructs which are borne out of their own set of cultural practices and linguistic ideologies. In other words, the desire of transcending one's own culture and language practices would have to be viewed as rooted in particular cultural practices, and as such can never become "objective" or in its absolute sense "culture-independent".

    Not withstanding Wierzbicka's on-target critique of the ethnocentric universalism of traditional emotion theories (cf. also Goddard in press), it is questionable whether the search for a universal understanding of emotions (or any concept or term) can contribute to a better understanding of what emotions mean to people who speak different languages and who "belong" to different cultures. 6 Unfortunately, it remains unclear how, to use her own terms, "language imposes its own classification upon human emotional experience" (1995a: 236), i.e. how the experiencer in actual concrete situations transforms "the culture-independent psychology of human cognition and emotion" (1995a: 233) into language- and culture-specific concepts of "how-to-think" and "how-to-feel", and how these more specific concepts turn themselves into situated emotion talk in which participants are held accountable and blame is attributed. In spite of the seeming challenge of figuring out the "real" set of semantic universals (one that might hold for all emotion terms in all languages), one is left wondering whether Wierzbicka's rather undertheorized view of the cognition-language relationship has anything to contribute to how people in actual discourse settings talk - particularly in situations when they feel the need to implicate their own (or somebody else's) feelings or emotions - overtly by use of emotion terms, or covertly by use of other linguistic (and nonlinguistic) means. Thus, I am not advocating in general to dismiss investigations of the lexicon of a language as unimportant, as Wierzbicka wants to understand my position (Wierzbicka in press). Rather, I consider lexical items in use for discursive purposes, as will be shown further below, as quite relevant for a discursive analysis. 7 Although not coined to apply to Wierzbicka's approach specifically, White's critical characterization of most semantic approaches to emotions holds water in this case as well; in essence, they all are "narrowly lexical (and ultimately limited by the absence of contextual and performative information) " (White 1990: 65).

    Harré's discursive approach to emotions takes up on some of the more interesting aspects of Stein's and Wierzbicka's proposals and tries to carry them one step further. In keeping with the anthropological, cultural model approaches, Harré's emotionology adds the directive, cultural force of emotions to the representational system, i.e. the information about structure of experience (= shared knowledge). This opens up space for investigations of the processes through which cultural knowledge obtains motivational force for individuals. However, the knowledge of scripts is not viewed to be organized in terms of taxonomic structures, as in the early days of cognitive anthropology, and neither is culture a monolithic unit. Cultural meanings are potentially conflicting and more loosely organized (Holland & Quinn 1987). At the same time, Harré's leaning to view emotions (as all other psychological processes) as products of discourse, constituted in interaction, forces a new emphasis on the study of emotions which is more thoroughly discourse oriented. With regard to the question of how emotions are acquired, we are pushed to look for a form of cultural learning that can incorporate aspects of Stein's and Wierzbicka's cognitive approaches under the auspices of a discursive orientation.

2. A "linguistic-constructionist" approach to "emotions"

2.1. References to emotions as indexes

Before presenting some of my own research on emotion talk, let me reiterate the constraining assumptions for my own approach: My original interest was and is emotion talk, or more precisely, talk about emotions - or better: emotional situations. Traditionally, we do not employ talk about topics such as rock formations or thunderstorms to investigate or explore such (natural) phenomena in themselves. Only if we try to explore what experiences with (natural) objects mean to common (though at times also: uncommon) people, we may be interested in people's personal accounts. Thus, one of my guiding assumptions for my own investigations is that references to emotions are indexes not necessarily leading directly to the phenomenon. Rather, what is indexed is how a person wants to be understood. Thus, a term does not directly display its meaning, and an account of a happy or sad event does not directly display what happiness or sadness 'means'. The person - so to speak - "interferes" by 'wanting to be understood'. And as such, the same (emotion) term (or the same account of 'the same' event) might mean something quite different in different contexts; and similarly, in particular contexts, other language forms might have the same "meaning" as the (emotion) term (or the whole account) under consideration. Consequently, exploring the range of possible meanings of emotion terms - in the sense of what they are used for in emotion talk - is at the core of the following investigations. 8

    It also should be stressed that because of space limitations, this section cannot present the underlying approach to the relationship between language and emotion in its full scope. Rather, the examples and illustrations from some of my ongoing investigations into emotion talk in adults and children (and across different languages and cultures) are supposed to stand in for, and as such exemplify the orientation toward a more coherent approach to this very relationship, one that is meant to productively connect with the three approaches reviewed in the first section of this paper.

2.2. References to emotions in third-person accounts

My interest in the use of references to emotions (and to other "inner states") originated within a broader study of children's and adults' abilities to tell a 24-paged picture book (Frog, Where Are You?). The main task in this study consisted of creating the (pictorially presented) characters linguistically, and relating them in terms of their actions across time and space in the form of a cohesive/coherent narrative. One of the first insights regarding the use of references to emotions that came out of this project (reported in Bamberg 1987, and 1991) was that such references did not necessarily "originate" from the pictures: Narrators of the picture story - often - chose to override a pictorially presented facial expression of one of the characters with a reference to the "opposite" emotion. For instance, a boy, whose face was obviously expressing anger, and who was linguistically referred to as angry when the picture was presented as a single, isolated picture, was referred to as happy (by the same subject three minutes later) when referring to this picture in the narrating activity of establishing the Frog, Where Are You? story (see Bamberg 1991 for further discussion). 9

    References to emotions that were ascribed to the prot- or antagonists in the picture book narrations served two listener- orientations: First, in the sense that they seemed to refer to "internal states", they momentarily brought the flow of events to a halt. In doing so, these references marked the narrator's stepping out of the event line of the plot, and presented an evaluative stance or perspective with regard to the event under consideration. Second, these references typically occurred at episode boundaries, i.e. at the beginning or at the end of event sequences that were presented in a linguistically bounded fashion. In this way references to emotions served the function to "transfix" two conjoined episodes, thereby contributing to the episodic flow of the narrative whole.

    From these observations I concluded that references to emotions in this kind of narrative activity are not referential in the same sense as establishing a character or temporal and spatial reference points in narrative discourse. Rather, they frame narrative units (episodes), and in this function "they are pervasive qualifications of the events they span and inform" (Young 1987: 23). They signal to the listener how the different narrative units are connected, and in doing so, they reveal an "overarching perspective" from which the narrative whole is being constructed. 10

    Another insight came from minute comparisons of form-function devices, the way they were employed for the picture book telling task across different age groups of children and across different languages (Berman & Slobin 1994): When comparing the use of references marking the protagonist's "inner state" (i.e. his reaction to the appearance of one of the antagonists) as surprise and/or scare, it became apparent that some narrators made the same point by qualifying the action of the antagonist in this same scene as having occurred suddenly or unexpectedly. Thus, what at first sight looked to be a description of an internal state of the protagonist turned out on closer scrutiny to be the expression of a particular perspective for the discursive purpose of narrating. Characterizing actions or events as occurring suddenly is an instruction to view these actions or events from the point of the character to whom they come "suddenly" or "unexpectedly", and thus can result in a "surprise", or a "scare". Instructing the listener to take this particular point of view reveals the overarching narrative perspective from which the narrator has delineated single events and orchestrates them into a narrative whole.

    In yet another, separate investigation (Bamberg & Damrad-Frye 1991), we could establish that children learn to use these kinds of devices to first mark more local level narrative scenes, before they establish the functions of these devices for more global narrative perspective taking purposes. More specifically, at around the age of 4 years, children use emotion references to locally connect a single (precipitating) story event to an "internal" outcome "in" one of the story characters. At around 8 years, children begin to tie together emotions, motivations and story events from a more global perspective, orienting their listeners more clearly to the narrative whole. At all times, however, references to emotions (here as ascriptions to others) function to construct a particular perspective that links or transfixes actions that would otherwise be seen as unmotivated and therefore as unconnected. 11

    In sum, my early investigations of how references to emotions in picture book narrations were put to use led us to see these references as rhetorical devices that orient an audience toward a perspective from which characters are orchestrated in relationship to one another. At this point, the critique could be launched that this function is typical but specific to discourse about third- persons, of which the picture book narration is (yet) another specific case. Further, it could be argued that the narrator of these kinds of picture-books is forced to present the emotions of the third-person character from his/her (the narrator's) own perspective, because he/she has no direct "knowledge" (=privileged access to) a third-person's emotions. Thus, in order to decide more conclusively whether the established audience function of emotion terms is unique for accounts from a third-person vantage point, we turned to emotion talk that was conducted from a first- person point of view. 12

2.3. References to emotions in first-person accounts

In the following I will draw on an investigation in which we asked American-English-speaking children to tell emotion experiences from two different perspectives. More precisely, we asked 80 children ranging from preschool to 3rd grade (4-10 years of age) to tell us about "one time when they were angry/sad/scared/happy", and "one time when they made someone else angry/sad/scared/happy". Thus, with both elicitation questions children were required to "report" concrete, personally experienced incidents of so-called "emotion experiences". In response to the first question, they were supposed to present the experience from the perspective of the "I" as undergoer (where "the other" is to be constructed as the causal agent), while the second question asked them to place the "I" in the role of the causal agent who instigated the emotion experience (leaving "the other" to be constructed as the undergoer).

    The first finding of this investigation (reported in detail in Bamberg in press a, in press b) consisted of two quite different profiles in the responses to the two different elicitation questions: Presentations of I as causal agent [and other as undergoer] for anger-scenarios were typically done by construing the I as inagentive, and the other as vague and de-individualized (e.g. by way of pluralizing "them"). Further, the whole incident was presented in terms of a probable appearance (by use of modality markers such as could, might, probably, or maybe) or as a plain accident. Examples (1) through (3) illustrate such construals:

(1)    it was a couple of years ago

    when I took the crab away from my brother

    then I stuck my fist out

    and he ran right into it

    and got a bloody nose

(2)    we were fighting maybe

    I don't really know

(3)    I lifted the hockey stick probably

    and it hit them

    kind of by accident

    In sharp contrast, constructions of the other in the role of causal agent [and of the I as undergoer] in anger scenarios were achieved by individualizing the causal agent, and marking his/her actions as highly transitive and affecting the I:

(4)    I was in the room

    and my sister kicked me

    and it went right into the rib bone

    and I went down to my mother

    and told her

    my sister got into trouble

(5)    when my sister slapped me across the face

    just because she didn't let me in her room

The syntactic frame of these constructions followed a highly regular pattern: The other was kept in subject position, while the I occupied the direct object slot ("me"). Often implicit, though not plainly expressed in these constructions were implications that the act was not justified, such as in line 2 of example (5).

    In general, children seemed to organize the latter type of accounts in sharp contradistinction to the first type of accounts: While the other as undergoer was constructed as inagentive, the I as undergoer was highly agentive; and while the other as causal agent was constructed as highly agentive, the I as causal agent was highly inagentive. In spite of the fact that the scenario was kept the same ("someone does something that causes someone else to become angry"), it seems to make a real difference for (American) children as to who is doing what to whom.

    Of course, it should be clear that the issue in these two different construction types is the difference in discourse purpose: Construing the other as highly agentive when the I is the undergoer serves the purpose of attributing blame. The discursive force of this construction type is to align the audience with the person who gives the account, and potentially assist in a possible revenge scheme. In contrast, construing the I as inagentive when the other is the undergoer serves the purpose of saving face: Having been caught in the (narrative) act of inflicting physical harm on someone else is viewed as less aggravating when it was not fully intended, or at least, when the reasons for "who is to blame" cannot be clearly located.

    To summarize, accounts that supposedly report one's own feelings and emotions (as caused by others) or that report someone else's emotions (as caused by the same person who is doing the reporting) are fashioned for different discursive purposes. Grounded in these different purposes, the reports themselves gain their specific linguistic structure; none of them is "more real", "more true" or "less constructed" than the others. The way other and I are linguistically positioned with regard to one another at the plane of character construction cannot be viewed any longer as the linguistic re-presentation of events the way they "happened". Rather, these so- called "content arrangements" are re-presenting (=they stand in for) the way the speaker positions him-/herself with regard to the audience. 13 Thus, constructions of a textual reality are always co-constructed in concert with a local moral order for the purpose of "becoming understood".

    When these first-person accounts given in response to the two types of elicitation questions were compared to accounts in which the children had to view the same two emotion scenarios (of making someone else angry/sad/scared/happy and being angry/sad/scared/happy) from the point of view of a generalized other 14 , a different picture emerged: The differences in positions with regard to how the characters were aligned as well as with regard to the local moral order vis-a-vis the audience were washed out. This finding, which did not come unexpectedly, nevertheless underscored the relevance of the notion of discourse perspective, and its dependence on what general perspective is chosen by the narrator/speaker: The discursive purposes that hold for first-person accounts are clearly different from those that require the narrator to take the perspective of a generalized other, with third-person accounts leaving open different perspectives to chose from.

2.4. Talk about having two emotions at the same time

In a different study (reported in Bamberg, Ammirati & Shea 1995), we investigated children's and young adults' accounts of "having two emotions" simultaneously. Again, we elicited their accounts from three different actor-perspectives (first-person, third-person, and generalized other perspective), though here I will only touch upon their first-person accounts and summarize the more general conclusions and insights that we drew from this study.

    The following three accounts are typical answers to the elicitating question "Can you tell me about a time when you felt both sad and angry (examples 6 and 8) [happy and sad (example 7)] at the same time?":

(6)    I was sad that someone I love is leaving

    and I am angry that he left

(7)    I was sad that he [my grandfather] was dead

    and I was angry that he died

(8)    I am happy that I am going to college

    but sad that I am leaving my friends and family

First, it should be mentioned that children prior to the age of 7 face difficulties in fitting the two different emotion perspectives together into a coherent account, especially when it comes to two emotions of different valence (such as happy and sad). Older children, and, more typically, young adults, who generally are more apt to construct dual emotion accounts, nevertheless seem to find it more difficult to coordinate two "simultaneous" perspectives on "anger" and "fear", which are actually two emotions of the same valence.

    More relevant for the present purpose, examples 6 through 8 document clearly how narrators are employing linguistic means (construction types) to bring about the framing (or "illusion"?) of two simultaneous (emotion) experiences in one event construal: Apart from the different temporal reference points ("was sad" versus "am angry"), the narrator in example 6 employs the aspectual unboundedness of leaving to contrast with the aspectual boundedness of left, resulting in two different vantage points from which the same happening is being "perspectivized". The same perspectivization is constructed in example 7: The state description of "being dead" is contrasted with the process of an activity that linguistically construes a syntactic subject with its semantic role of a potential agent. In both examples, these differences in construction types impact on how the speakers seem to index their stance with regard to agency and responsibility: States and agentless happenings typically result in "sadness", with no animate agent to blame, while situations that evoke "anger" are more likely the results of willful, intentional actions brought about by animate others.

    Example 8 illustrates a similar technique of event construal, here by use of two different perspectives on the situation of transitioning from high-school to college. While both predicates construct this situation by use of motion verbs, leaving focuses on the source, keeping the telos unspecified, and thereby orienting the audience with regard to the transition from home to college in a backward fashion. In contrast, going keeps the source unspecified, and focuses on the telos of the motion. In this way, this construction orients forward, toward a future reference point. Simultaneously, the telic orientation of going in this example construes a more agentive perspective for the transition event, thereby setting up the contrast to the less agentive perspective for the act of leaving. Though the contrast between these two construction types does not foreground as clearly as in the previous two examples who is responsible and blameworthy (as in "anger"), and what happened agentlessly and by "accident" (for "sadness"), the leaving perspective chosen for the event construal in example 8 nevertheless resembles the inagentive construction type that is typical for sadness accounts from third- and first-person perspectives.

    The discussion of the examples is oriented to extrapolate two related points, one holding for the assumption of "having two emotions simultaneously", the second for "having emotions" in general. The first is meant to address the production of the appearance - or, if one prefers: illusion - that is achieved by these particular linguistic construction types. The accounts given for situations of two simultaneously existing emotions are linguistically constructed from two different perspectives, perspectives that are not existent or available a priori and outside of language. In other words, the ability to linguistically take different perspectives (for different discursive purposes!) on the same "event" 15 is a prerequisite for the construal of this event type called dual or mixed emotions (for which English does not provide a class of lexical items). Or, put more strongly, this discourse ability to take different perspectives on events, and to hold these perspectives concurrently, is the basic constituent out of which conceptual impressions such as "having two emotions simultaneously" are made of. Consequently, this linguistic ability cannot be seen as being derived from two "actual" feeling constellations that are simultaneously "experienced", such as the simultaneous feelings of anger and sadness. Rather, these so-called "feeling states" are the products of conceptual framing, which in turn are the products of the linguistic ability to take perspectives. If we accept that "having two emotions simultaneously" cannot be derived from the composite of two concurrent feeling states, or from the overlap of two emotion concepts (applied to one situation concept), but that it is the product of the linguistic ability to view a situation for two discursive purposes, we may be forced to apply this insight also to how references to single emotions are maintained and achieved. We will follow this argument in a more elaborated fashion in the concluding section (section 3). However, let me recall briefly here that Trabasso and Stein (1993), who came to the same conviction, namely that reports of so-called double-emotions (two simultaneously occurring feeling states) are (retrospective) accounts of what might have actually taken place sequentially (rather than at the same time), argue along similar lines, namely that it is the verbal account that 'gives rise to the illusion' of the actual possibility of holding more than one emotion at the same time (see endnote 4 above).

2.5. Confusing "sad" and "angry" - a case for genre in the appropriation of emotions

2.5.1. Introduction to the study In this section I want to present in more detail than in the previous sections a study in which children's responses to the two questions of "being sad" and "being angry" are compared across the four age groups. While primarily being concerned with accounts given in the first person (the genre of personal, past experiences), we will consider some features of accounts of the same situations given in the generalized person perspective (the genre of explanatory discourse). To briefly recall the elicitation conditions, the first person accounts were responses to the question: "Can you tell me about one time, when you were angry / sad? ", while the generalized person perspective was supposed to be adopted by answering the question: "Imagine I am from far far away, and I wouldn't know what it means to be angry / sad. How would you explain to me what it means to be angry / sad?"

    The aim of laying out this study in considerably more detail is threefold. First, it really seems to be of importance to convey a sense of methodology, i.e. that the investigation of the relationship between emotion, cognition, and language is grounded differently and follows a different line of inquiry. Thus, the methodological pursuit in the studies summarized thus far and presented in more detail next is not simply focusing on a different "level of analysis" when compared with the frameworks of Wierzbicka and Stein presented in section 1 above. Rather, it starts from some rather different assumptions and leads to different insights, some of them compatible, while others may be not. Second, it is deemed to be of utmost importance to give the reader some better insights into the actual nature of how constructionism "constructs" the building blocks out of which social and individual constructs are made. Thus, in the following we will specify in considerably more detail what processes are at work when emotions are given meaning. In particular, we will turn to linguistic, grammatical analyses to show how intricate the building blocks are interwoven and embedded in social practices. Third, the following study is explicitly developmental, and therefore in its very nature can reveal interesting insights into how the child (here the American English-speaking) puts together language, thought, and emotion in the act of communication. This, in turn, can lead to improve our theoretical considerations with regard to the relationship between language, cognition, and emotion.

    Actually, this particular investigation originally was stimulated by an accidental, though intriguing observation that we stumbled across in the process of data elicitation: When asking the children ranging from preschoolers to 3rd-graders a battery of 12 consecutive questions, all revolving around 'having been angry, sad, scared, and happy', we repeatedly heard some of the children maintaining that one or another question had already been answered. Looking closer at where this happened, we realized that these kinds of confusions occurred solely with questions regarding angry and sad, but never with any of the other ones. In addition, when we asked these children to give us an answer anyway, most often, the accounts given were word-for-word repetitions. At a later point in the investigation, when we decided to go back into the actual recordings, in order to obtain a count for how many times these comments from our subjects were made, we realized that these "confusions" were more typical for the younger children, and accounted for more than 15% of the answers of the preschoolers and kindergarteners. Furthermore, it appeared as if these "confusions" only occurred in response to the first-person questions: "Can you tell me about a time when you were angry/sad?"; rarely in response to the third-person questions: "Here is a little child whose name is XXX. One day s/he was very very angry/sad. Can you tell me what happened that day?"; and never in the generalized person perspective as a response to the questions: "Can you explain to me what it means to be angry/sad?". These observations, although based on very preliminary evidence, nevertheless seemed to us intriguing enough to follow up with the investigation into the question of whether or not there is actually any support for the assumption that (American English-speaking) children might at an early age confuse the two emotions anger and sadness. And if this turns out to be the case, it should be determined, whether this confusion is more typical for any of the different genres under consideration.

    Considering first Stein's findings of children's abilities to reason about angry and sad experiences, such "confusions" should not have occurred, because "children as young as 3 years of age are able to perform at a high level of proficiency" in figuring out the different components that lead to emotional states and that follow from them (Stein & Trabasso 1992: 241). However, Stein and her associates also point to the similarities between anger and sadness in that both emotions are typically evoked by the same goal outcome patterns, i.e. both are typical reactions to a loss or aversive state (Stein, Trabasso & Liwag 1992: 291ff.) 16 , and therefore, with development, it is more likely that both can be experienced as results to (though different) appraisals of the same event (see, however, with regard to whether the two emotions are actually experienced simultaneously or sequentially, Trabasso & Stein's (1993) cautious remarks, as well as the discussions in the previous section regarding "double emotions").

    With regard to other sources that report a potential fusion (or confusion) of anger and sadness, I could only locate some spattered hints in linguistic and anthropological reports about a number of African languages which, at least at the lexical level, do not seem to draw a distinctive line between what in English is divided into anger and sadness. Drawing on reports by Davitz (1969) and Leff (1973) as their sources, Heelas (1986) and Matsumoto (1994) maintain a similar opinion, namely that "some African languages have one word that covers what the English language suggests are two emotions - anger and sadness (Leff, 1973). Likewise, Lutz (1980) suggests that the Ifaluk word song can be described sometimes as anger and sometimes as sadness" (Matsumuto 1994: 120). In addition, Heelas argues that "English- speaking Ugandans do not distinguish between 'sadness' and 'anger', as we do (crying being an important feature of our distinction but not for Ugandans)" (1986: 239). It remains unclear from the reports, what aspects actually are expressed in Swahili, and what supposedly might lead bilingual Swahili speakers to use only one English lexical expression for covering both meanings of English sad and anger. Actually, the literature reported does not even say which English expression is used, angry or sad. We found these reports nevertheless intriguing, because they can be taken as pointing toward potentially different arrangements that take place lexically and/or conceptually when it comes to how emotions - here anger and sadness - are made sense of.

    Another interesting observation that points in the same direction stems from Fischer's (1991) investigation of how Dutch speakers make sense of the Dutch equivalent to 'anger'. In evaluations of the emotion 'anger', Dutch adult subjects displayed two different attitudes according to where the incident that led to the emotion took place: "In private situations respondents do not wish to hurt others, nor gain contradictory effects in the behavior of others. However, they want to show commitment to others by expressing their anger. In public situations, on the other hand, one is far more concerned with how others will evaluate one's anger and so the anger seems primarily to be used as device to maintain or improve one's position" (Fischer 1991: 151). This observation leads to an interesting speculation with regard to a differential in functionality of these two different evaluative stances: The expression of anger in public settings is more likely to be valued as negative, while its expression in private settings is argued to be geared toward some more positive ends. 17

    Taking these insights from Heelas' and Matsumoto's reports of some African languages' collapsing into one word what is differentiated in English by the two lexical items anger and sadness, and Fischer's discussion of the Dutch equivalent of 'anger' (which strongly resembles American attitudes toward anger), we are reminded of the possibility that there are a number of potentially "confusing" components in how to make sense of anger and sadness; or at least, that it may be a whole lot easier to confuse some of the components of anger and sadness than, let's say of happiness and anger. Considering further, that children in their language acquisition process have to learn to sort out these components, it may be very possible that they might go through phases of confusing certain components in certain situations. Thus, a closer look at the different kinds of situations where emotion talk is used (for potentially different purposes) might be exactly the route to travel to find out.

2.5.2. Subjects, Data Elicitation, and Coding Since we will report in this section the data for all four age groups, let me briefly summarize the subject population, the elicitation technique used, as well as the ways the data were transcribed and coded.

    Subjects The participants of the study reported in this section were 80 children from four grades with 20 children each: preschool, kindergarten, first graders, and third graders. Although all subjects were from the same regional area in Massachusetts, USA, the school populations varied. The preschool and kindergarten children attended primarily middle-class schools, whereas the first and third graders attended a racially and economically diverse school. The mean ages for each group were: preschool: 5;2, kindergarten: 6;1, first-grade: 7;3, and third- grade: 9;1 years, respectively. In order to be a participant in the study, the children had to meet four criteria: First, the child's parent had to give permission (as well as the child) at the time of the interview. Second, the child could not have any diagnosed language or learning impairments. Third, the children needed to be fluent speakers of English. Finally, the children had to finish the interview by giving answers to most of the questions.

    Data Elicitation The interviewers for this study were trained so that the children were tested using a consistent format. All interviews occurred at the schools but in separate areas away from the regular classrooms. Each session was tape-recorded for the purpose of later transcription. The interview consisted of 12 questions encompassing three genres on four topics: The genres consisted of (i) the first-person (personal / past experience) narrative, (ii) the third- person narrative, and (iii) the generalized-person (explanatory) account. The topics consisted of emotion situations: (a) being angry, (b) being sad, (c) being scared, and (d) being happy. Thus, the questions asked in the interview took the following forms: (i) Can you tell me about one time, when you were angry / sad / scared / happy?, (ii) Pretend you know a little {boy/girl} whose name is {Andy/Anna - Samuel/Sarah - Steve/Sally - Molly/Harry}. One day {he/she} was very very angry / sad / scared / happy. Can you tell me what happened that day?, and (iii) Pretend I am from far far away and I wouldn't know what it means to be angry / sad / scared / happy. How would you explain to me what it means to be angry / sad / scared / happy? All 12 questions were administered to each child in random order, with the only two restrictions to start each interview with genre (i), and to avoid asking two questions about the same topic in a row.

    Transcription and Coding Following the transcription format laid out by Berman, Slobin, Bamberg, Dromi, Marchman, Neeman, Renner and Sebastian (1986), all verbal responses were transcribed in clauses, giving each individual clause a new line. Since it had been mentioned in previous studies that anger and sadness differ particularly in terms of actor involvement (transitive versus intransitive acts) and type of action (intentional versus accidental, and justified versus unjust), we determined to develop categories that would tap those dimensions. Accordingly, the transcripts were coded along the following dimensions:

[A]    Behavioral symptoms (in particular: facial expressions):

        eg:    "in tears" or "crying"

[B]    Actions explicitly marked with intent of actor versus without intent

        eg:    "I told her to stop but she didn't" (= with intent)

            versus:

             "by accident" or "but she didn't really mean it" (= without intent)

[C]    Actions explicitly marked as unjustified versus justified:

        eg:    "but I deserved that" (= just)

            versus:

            "I broke my sister's arm, and mom said I was grounded for life"

                        (= unjust)

            versus:

            "my grandma died" (= just or unjust not marked or not at issue)

[D]    Transitivity of action (with an animate actor in subject position):

        eg:    "my brother bit me" (= high in transitivity)

            versus:

            "my sister moved away" (= medium in transitivity)

            versus:

            "my sister lives far away" (= low in transitivity)

In addition, we decided to apply the transitivity code also to clauses which had no actor in subject position (eg. "my glasses fell in the trash"). Although these clauses by their very nature rank lower in transitivity, particularly since they seem to place emphasis on states or results of actions (eg. "my glasses were in the trash"), we nevertheless considered it worthwhile to be able to compare the different genres along those lines to get an impression of what we termed "overall transitivity" (versus "actor transitivity").

    While the coding for category [A] turned out to be straight forward, equally high interrator reliability scores were maintained for categories [B] and [C] (89%). However, the rating for transitivity turned out to be much more difficult. In actuality, the medium category was the hardest to determine, while deciding whether clauses fell into the high versus the low transitivity category seemed to be more of a clear cut case. On the basis of this, we made the coding decision to use the medium category as little as possible. In other words, the raters were asked to enforce wherever possible a 'high' versus a 'low' ranking, and only in cases where there was absolutely no easy choice, the mid-level category was to be used. Using this method, we could attain an equally high interrator reliability score (89%).

    It should be noted that contrary to most introductory methods books, the codes applied in this as in most other studies are not strictly speaking hypotheses. Coding for "transitivity" originally (see Hopper & Thompson 1980) entailed a set of 10 parameters each ranging along a continuum from high to low transitivity. Here, it is simply aimed to reveal whether the 'sense of agency' that was conveyed in a particular unit was "high" or "low". In this sense then, the coding dimensions chosen are preliminary attempts to check whether there might be hidden more along the dimensions. What the mentioning of "behavioral symptoms" (in the form of tears for instance) or what the choice of low transitivity constructs actually mean, that is what they index in the communicative act, will become part of a more refined analysis in a subsequent analytic step to be reported in section 2.5.5.

2.5.3. First Findings Since it is the proclaimed purpose of this study to investigate the questions of potential differences along dimension (a) [= the topics of being angry versus being sad] and dimension (b) [= first versus generalized person genre], the remaining two topics (being scared and being happy) as well as the genre of third person narrative will be neglected in the following report. Chi-square analyses were performed on all variables to determine significant developmental changes. The percentage figures reported in Tables 2 through 4 (coding for dimensions A-C) are based on an n of 20 per age group. The figures in Tables 1 and 5 (dimension D) are based on occurrences of forms per overall clauses per age group. 18 Taking the four above coding dimensions as quasi hypotheses along which differences were expected, the following findings were revealed:

    Actor Transitivity Ratings in the First Person Genre As presented in Table 1, angry scenarios, in which an animate actor is mentioned, are clearly marked (i.e. linguistically constructed) in terms of highly transitive events. Low transitive actions are very rare. With regard to sad situations, the picture is not quite as clear: While sad situations rank lower in high transitivity than angry situations, and while they also rank higher in low transitivity than angry situations, they cannot be characterized in terms of a clear preference for low transitivity. However, there seem to be some developmental changes in the role of transitivity in sad scenarios: While younger children are more likely to construct them in terms of high transitivity, older children are more likely to construct them in terms of low transitivity. (Note that within this coding category only those clauses are coded that do mention an actor in subject position.)

    ------------------------------------------------------------------------------------------

            Insert Table 1 after the above paragraph

--------------------------------------------------------------------------------------------------------------------

    Behavioral Symptoms When it came to the mentioning of behavioral aspects of the emotion situation on the part of those who were characterized as affected, there was a clear distinction between the two emotions, holding for all age groups and for both genres: If tears or a particular facial expression was mentioned, it happened to characterize sad situations (see Table 2 for the exact figures). (Note that the percentage figures given are based on an n of 20.)

    ------------------------------------------------------------------------------------------

                Insert Table 2 about here

--------------------------------------------------------------------------------------------------------------------

    Explicitness of the Intentionality of the Action As can be seen from Table 3, sad scenarios are much more likely to be characterized in terms of an action that was carried out unintentionally. This observation holds more strongly (and more clearly) for the generalized person genre than for the first person genre. In addition, there seems to be a developmental change holding for the characterization of sad scenarios in the first person scenario in terms of unintentional actions, which does not hold for the generalized person scenario. (Note that the percentage figures given are based on an n of 20.)

    ------------------------------------------------------------------------------------------

                Insert Table 3 roughly here

--------------------------------------------------------------------------------------------------------------------

    Explicitly Unjust Actions Although there are a few children who explicitly state that the action that has led to a sad experience, was unjust, this nevertheless is a very small minority, and does not hold at all for generalized sad experiences. However, as can be seen from Table 4, to mention explicitly that an action was not justified, seems to be of high relevance for the construction of anger scenarios; though interestingly much more so for those scenarios in which anger was self-experienced, i.e. for the establishment of the first person perspective. (Note that the percentage figures given are based on an n of 20).

    ------------------------------------------------------------------------------------------

                Insert Table 4 roughly here

--------------------------------------------------------------------------------------------------------------------

    Overall Transitivity Ratings for Sad Scenarios Table 5 reveals a more thorough look at the changes that take place in characterizations of sad scenarios. Note that in contrast to the analysis presented in Table 1, all clauses were coded for high versus low transitivity, resulting in a more clear-cut system than in the "actor transitivity" analysis. While younger children clearly characterize sad situations in terms of high transitivity, this tendency seems to fade with increasing age. Along the same lines, with increasing age, children turn to a marking of sad situations in terms of low transitivity. In addition, although the same trend holds in general, there are nevertheless interesting differences in the way it surfaces in the two different genres: While high transitivity is a major salient characteristic for younger children in the first person genre, low transitivity becomes by far the more salient characteristic for the older children.

    ------------------------------------------------------------------------------------------

                Insert Table 5 around here

--------------------------------------------------------------------------------------------------------------------

2.5.4. Summary and Preliminary Conclusions Having used the four coding dimensions as preliminary grids according to which we will be able to assess similarities and differences between the two topics (of being angry and being sad) and the two genres under consideration, we can more clearly outline the following areas that need to be followed up on in a second more refined type of analysis. First, around the age of 9 years, as documented in the responses of the third graders, American English-speaking children seem to construct "being angry" and "being sad" by use of different linguistic means, employing constructions that revolve around the issue of transitivity. Being angry is constructed by use of high transitivity constructs, and being sad by use of low transitivity constructs (see Table 1). In addition, when it comes to letting the listener know aspects of why the action (that evoked the emotion) occurred, and what behavioral consequences those actions had in the affected person, older children clearly differentiate between anger and sadness: Actions leading to sadness are more likely to be unintentional actions (Table 3), and they often result in crying and tears (Table 2); while actions leading to anger are more likely to be unjust actions (Table 4).

    With regard to these latter more explicit markings of the intentionality and justification aspects of the actions and their behavioral, bodily consequences, it looks as if younger children as well 'know' how to differentiate between the two emotions, and use those markers accordingly. However, with regard to their use of transitivity constructions, a different developmental picture emerges. While the construction of angry situations does not undergo any developmental changes regarding transitivity constructions, sad situations are more likely to be marked by use of high transitive constructions by the younger children (Tables 1 and 5). Only the oldest age group of third graders clearly constructs being sad situations by use of low transitivity constructions.

    Turning to the differences between the two genres, we find a slightly higher employment of "crying" and "tears" (= behavioral symptoms) in explanations of sadness, and along the same lines more mentions of the actions as being unintended in explanations of being sad. Explicit mentions of the fact that the action that caused the anger was unjust, however, is much more typically employed in constructions of the first person genre, that is where the narrating self was presenting him-/herself as having been perpetrated. Although there are some fluctuations in the figures between the different age groups, it can be maintained that the general distinction that is achieved by these constructions is already in place at an early age in the preschoolers. Again, when we turn to how high transitivity versus low transitivity constructions are employed to demarcate between anger and sadness, we find a quite different picture. Apart from the developmental shift in the construction of sadness, as reported in the previous paragraph, high transitivity markings clearly prevail in the first person genre, while low transitivity constructions clearly prevail in the explanatory genre (see Table 5).

    We could stop here and conclude that American English-speaking children at a relatively early age, i.e. around the age of 5 years, already clearly distinguish between the two emotions anger and sadness, particularly when it comes to how the actions leading to the emotions are to be evaluated (intended/unintended and just/unjust) and what kinds of behavioral consequences these actions cause as a result. Only when it comes to the more fine grained level of the English language use, young children have to do some "sorting out", in the direction suggested by this study; though this further, linguistic, level of differentiation and sophistication is not really relevant to the conceptual distinction between the two emotions, what they are, what they mean, and how they function in the interpersonal regulation of relationships. This distinction, it could further be maintained, is already established in its most basic form. And the remaining bits of confusion are due to children's linguistic organization, not to any categorial or conceptual issue.

    However, this conclusion not only would fall short in accounting for what we set out to explain, namely the original confusions in having to give sad accounts versus angry accounts, where young American English-speaking children seemed to take both situations to mean the same, but it also would fall back into the peculiar segregation of language from cognition that we were trying to overcome (cf. section 1.4. above). In addition, the differences in children's use of construction types in the two different genres seem to reveal interesting clues with regard to what is involved in the process of constructing sad situations early on in the same way as angry situations, and how this early confusion is resolved gradually in the course toward a fuller and better integrated understanding of anger and sadness. To accomplish those aims, a second analysis was performed on the same data, one that focused more strongly on the actual constructions that were employed in the responses of the children.

2.5.5. The Grammars of 'Being Angry' and 'Being Sad' and Their Development. The term GRAMMAR in this context is not used metaphorically, to imply that emotions have or are a set of rules to be followed just like the grammars of languages. In contrast to this mechanistic view of language (or emotions), I want to make use of the term grammar as the strategic employment of linguistic constructions. Linguistic constructions which are not just 'rhetoric', not just employed to be better understood or to enhance intersubjectivity or the like. Rather, grammar in this context consists of a set of constructions which are instructions, i.e. they are the signposts for the audience to capture the directive (or a set of directives) that is emitted. Thus, grammar as a set of linguistic constructions is directly tied to the discursive purpose to which the particular constructions are put to use. As such, the terms 'grammar of being angry' and 'grammar of being sad' are attempts to capture the discursive organization of such topics as 'anger' or 'sadness'. The way the analysis will proceed in this section resembles in many ways that of Capps and Ochs (1995a, 1995b), whose research focused on the discursive organization of 'helplessness' and 'abnormality.'

    The Grammar of Being Angry in the First Person Genre. As already touched upon above (section 2.3 and Table 1 in the previous section), children of all age groups (as well as adults) construct angry situations relatively consistently in terms of high agency on the perpetrator's part. Examples (4) and (5), repeated below, document this type of construction:

(4)    I was in the room

    and my sister kicked me

    and it went right into the rib bone

    and I went down to my mother

    and told her

    my sister got into trouble

(5)    when my sister slapped me across the face

    just because she didn't let me in her room

    and I wanted to play a game

    but she didn't let me

    and she slapped me across the face

The linguistic devices characteristic of the grammar of American-English anger can be summarized as consisting of a conglomeration of: (i) a highly individuated agent ("my sister"), and a highly individuated undergoer ("me"), (ii) marking the action as highly transitive (often elaborating on visible or otherwise empirical effects of the action), (iii) positioning the "I" as the experiencer and target of the action in the direct object slot, and (iv) the "other" (the agent) in subject slot. These aspects of the grammar of anger overlap with how linguistically two discursive purposes are achieved. On the one hand, the construction of a highly individuated target of others' actions may result in some form of empathy or sympathy on the part of the listener in the act of communicating. On the other hand, by emphasizing the "other" as the topical focus in the position of the syntactic subject, blame is attributed to this person. And this topical focus, so to speak, overshadows the discursive purpose of empathy. The topical construction can be taken to subordinate the empathy goal to the blaming purpose: the more empathy, the better the blame. The empathy aspect is subordinated and as such directed toward the topical focus, i.e. the "other" as the perpetrator.

    The Grammar of Being Sad in the First Person Genre. Typically in the accounts of older children (as well as adults) American sadness is constructed by use of two possible options: (i) positioning the "other" in subject position, as in example (9), accomplishing the construct of an inagentive happening,

(9)    it was when I was about 5- or 4-years old

    my biggest sister got into a car accident

    so she died

    because of a car accident

    and I was really sad for a few weeks

or by use of strategy (ii), positioning the "I" in subject slot, accomplishing the construct of an undesired state, as in example (10):

(10)    I was in Charlton

    and I moved to Worcester

    and I couldn't see my neighbors and their dogs

While strategy (i) entertains the possibility to make the other (here my sister) the potential topical focus, and as such keeps open the possibility of discursive blame, this possibility is systematically ruled out by two concomitant linguistic devices, (iii) the avoidance or downplay of the other's agentivity, and (iv) the lack of a target or the lack of any directedness of the action or activity. Those two devices perform a similar function on strategy (ii), denying the "I" to achieve the linguistic status of a topical focus, with blaming oneself as the discursive orientation (as in example 10, where the narrator could have continued by making his decision to move the topic, lamenting about it or his decision making power, which, however, is not very likely for a 9-year old boy).

    Thus, it can be maintained that the grammar of American-English sadness differs from the grammar of American-English anger in terms of complexity. While anger comprises the two discursive aspects empathy and blame, and as such requires a delicate balance between these two purposes, sadness seems to be more simple, because it is geared 'only' toward the discourse purpose of empathy. However, taking the prototypical English sentence structure of the transitive scene, the grammar that is fashioned to entertain the discourse purpose empathy is more complex, since it is a deviation from the prototype: It requires the downplay of the topical focus. Or, in other words, after a subject has been established which typically serves as the topical focus ("my sister" in example [9], and "I" in example [10]), this focus has to be "defocused" in subsequent clauses, as is done in example (9) by the explanation "because of a car accident", i.e. "she didn't really DO it herself". Thus, although the construction of anger is more complex with regard to its discursive functions the construction of sadness - in English - is more complex with regard to its linguistic construction.

    Turning to the constructions of sadness in the first person genre in the younger children, we often find accounts which strongly resemble their anger accounts in as far as the same linguistic constructs are employed. Examples (11) and (12) illustrate this construction type.

(11)    when Nikki hit me in the eye

    I was really really sad

    I cried for a whole half an hour

(12)    my Mommy hit me

    she hit me in the eye

    and I was sad

    and cried

It should be noted that the behavioral symptom of tears or crying in many of the younger children's answers is preceded by the declaration of the inner state and the action that has led to that inner state. In terms of the sequential arrangement this behavioral symptom is not only consequential, but its arrangement simultaneously communicates that nothing else was done. As such it contributes to an interpretation of 'helplessness', and, since the role of a comforting other is left vacant, it can be taken as an indirect request for comfort or empathy. Thus, the younger children's responses typically consist of two components, the first topicalizes the perpetrator, and constructs a highly transitive event, which is likely to be taken by the listener as an act of blaming; the second component, however, shifts from the "other" to the "I" as the new topic, resulting in the likelihood to be taken as the communicative attempt to elicit empathy.

    Within the frame of these structural characteristics, the younger children's constructions of sad scenarios (in the first person genre) are different from the older children's in that they consist of two units which co-exist next to one another. While the older children either topicalized the "I" such as in example (10), or they shifted from an inagentive "other" to a sad "I" in subject position, the younger children seem to be stuck with a highly agentive "other" in subject position as the topic of what the scenario "is all about". Thus, the obstacle of changing the focus from the highly agentive "other" to the helpless "I" is too difficult to overcome, and the attempt to construct a sadness scenario ends up very much like the construction of an anger scenario. Before turning to the question how American English-speaking children learn to differentiate successfully between being angry and being sad along the lines laid out in the personal accounts of the older children, let us briefly consider how anger and sadness are constructed in the explanatory genre.

    The Grammar of Anger in the Explanatory Genre. The construction of anger in explanatory discourse is achieved by a number of construction types. (i) An unspecified agent in subject slot, most often plural "they"; (ii) an unspecified target of an activity in direct object slot, most typically "you", where it remains unclear whether you refers to an unspecified hypothetical person or to the interviewer; (iii) an activity verb which nevertheless is much less specific (and therefore less transitive) when compared to the verbs used typically in the first person genre (eg. "doing something" or "hurting" in the explanatory genre versus "hitting" and kicking" in the first person genre); (iv) the clause modus is most often marked by "if" or "when", in conjunction with (v) the present tense which takes the situation into the realm of a possible world. The following two examples display the syntactic frames used to construct the generalized person perspective by use of which "being angry" is 'explained':

(13)    if someone hurts you

    and you get really really really mad

    then you are angry

(14)    you are angry at someone

    because they did something to you

    and you didn't like

    what they did 19

In more general terms, these devices lead to the overall construct of a being angry situation that is less of a bounded event, less vivid, and presented as much more detached than the same situation in the first person genre. Seeking empathy or blaming anyone for any wrong doing clearly don't matter. If there is a discursive purpose, it lies in 'describing' or making a general activity explicit, though from a detached angle. The listener is taken 'out there' into the world of generalities, seeking disengagement and estrangement from the realm of subjective experience; whereas in the first person genre the listener was led into the 'here and now' of subjective experience, to take part in its consequences, enjoying the mutual engagement of seeking blame and/or empathy.

    Children of the younger age groups gave shorter situation descriptions than the older ones such as "somebody hitting you". Almost half of the preschoolers, in one or another way, tried to weave into the situation a first person perspective such as in the following two examples:

(15)    see when Mommy always gives me a shower

    I get very angry

(16)    if somebody did something wrong

    if somebody chewed someone's candy

    then I could be angry

Even after having given a description from a detached, generalized person perspective, the child in the following example still seems to feel the need to give an exemplar in the first person genre to clarify his answer:

(17)    like if somebody kept on hitting you

    and you said 'stop stop'

    and they didn't

    and that's what also happens to me

    cause my brother ...

As these examples document, the younger children do not systematically use the present tense to signal the atemporal, hypothetical character of the situation, nor do they systematically use the simple past in order to characterize the past, though potentially normative quality of the incident reported. Tense choices as well as modality markers (can, could, would) are tokens for the younger children's efforts in trying to sort out the kind of detached stance that is asked for in the explanatory genre.

    The Grammar of Sadness in the Explanatory Genre. As was revealed in a preliminary way in Tables 1 and 5, all children constructed being sad explanations by use of far more low transitivity constructs (and much less high transitivity constructs) than in their first person accounts. Similar to the accounts given in the first person genre, a happening is constructed in which either "something is happening to you", as in the following example, or "you want something but you can't get it". Only at two occasions were actions of others constructed as leading up to a situation of being sad; one of them in the next example, where, however, the agency of those others was somewhat transformed by the pronoun "it":

(18)    when like someone calls you four-eyed

    if you have glasses

    you get not mad at them

    but it hurts your feelings

    and you're sad

    In contrast to the older children, younger children often considered behavioral displays of being sad as sufficient to describe, and, presumably, explain "what it means to be sad". "Crying" or "if somebody cried" were such typical answers. The only two occasions in which transitive actions of others were depicted as leading up to sadness, however, were explanations that were incorporating components strongly resembling the first person genre:

(19)    I would be sad

    because James hit me

(20)    when Nikki fighted with me

    like he did to me yesterday

    I cried a lot

More common is the construction of a happening scenario that focuses on a desired object or state one is not able to have as in the following two examples:

(21)    if you couldn't go to kindergarten

    like Steven

    he couldn't graduate

(22)    like your favorite blanket was up high

    where you couldn't go

Thus, in spite of some difficulties in sorting out the linguistic complexities of the explanatory genre at an early age, all children clearly demarcate in the explanatory genre what it means to be angry from what it means to be sad, construing the former as a bounded, though generalized and somewhat detached event, and the latter as an inagentive happening. The discursive functions of attributing blame or eliciting empathy are backgrounded in light of the discursive purpose of 'being descriptive'. 20

2.5.6. Appropriating Emotions: The Developmental Process of Disentangling the Grammars of Anger and Sadness. In an attempt to pull together the insights gained in the above analyses, we are now better equipped for locating the origins for children's early confusions between angry and sad scenarios and for delineating some of the factors involved in the developmental process between preschool age and the time children reach third grade. First, the evidence assembled clearly points toward the early constructions of sadness accounts in the first person genre as the "source" for the confusion. In the attempt to determine how these accounts differ from later sadness accounts, and also how they differ from accounts of anger experiences(across all age groups), we realized that anger accounts typically consist of the construction of a highly agentive "other" who is introduced as the discourse topic. This construction is in direct service of the discursive act of attributing blame. Since in English this discursive strategy is achieved by use of the prototypical sentence format that endorses "the transitivity scene" (Budwig 1995, Hopper & Thompson 1980), the construction of anger scenarios are grammatically relatively easy. Sadness scenarios, however, require a deviation from this more prototypical syntactic format. If another person has been introduced in subject slot, and therefore is likely to be taken to be the topic of the account, the narrator has to de-emphasize this person's agency in order to avoid the invocation of blame. He/she needs to re-orient the listener to an empathetic stance towards the person who gives the account. It is exactly this problem of re-orientation which younger English- speaking children face in their accounts of their own sadness experiences.

    Considering that the generalized person perspective is linguistically more complex than the first person perspective, and that the younger of the children struggled considerably in coming to grips with the timeless as-if modality of this genre, it should come as a surprise that anger and sadness accounts are relatively clearly differentiated in the explanatory discourse genre along the dimension of transitivity/agency. Thus, the confusion in the younger children between being angry and being sad cannot be traced to the general unreadiness of linguistically presenting what has conceptually already been mapped out. Rather, the early underdifferentiation between the two types of accounts lies clearly rooted in the pragmatics of emotion talk, more specifically, in the respective discourse purposes of attributing blame and eliciting empathy.

    With regard to what it is that pulls the 'confused' child 21 out of the state of underdifferentiation, and pushes toward a higher level of differentiation (and as such also to a higher level of integration), we have no evidence to say for sure. However, the way we were able to map out the developmental route from a clear state of underdifferentiation to a higher level of making sense of sadness. i.e. as a process of appropriating the tools necessary to talk meaningfully about the social relationships in which emotions are embedded, points up some highly important underpinnings. First, it shows that modeling emotional development in terms of an 'internalization process' of learning how to feel may not be sufficient. Further, we were able to draw out the limitations of modeling emotional development in strictly cognitive terms. As shown by this study, grammar, if understood correctly, i.e. not as abstract principles of a universalist nature, but as social know-how relevant for the construction of social meaning, plays an integral role in coming to grips with what emotions do and what they are used for in social communicative practices. 22 As such, learning to use the linguistic construction procedures for socially appropriate purposes is deeply embedded in cultural practices. To view this process in terms of appropriation rather than internalization gives space for the dialectics involved in the developmental process in general: On the one hand, the grammatical means - so-to-speak - are preformed. They have their social existence before they are put to use in social practices. It is for exactly this reason that children (or others) can come to use these tools inappropriately such as in their early sadness constructions in the first person genre. However, they are not predetermining their use apart and independent from their users. The child early on is practicing in a relatively autonomous way with these tools, assembling new construction parts with others that are already successfully in place. Thus, viewing this process of appropriating linguistic constructions in the determination of emotion meanings as an integral part of learning 'the language' adds an extremely relevant component to emotional development, probably one that is much more central than we were able to imagine thus far.

     In addition, and here we admittedly enter more speculative territory, having documented that the differentiation between anger and sad accounts took place developmentally prior in the generalized person, explanatory genre, before it could be appropriated in the first person, past experience narrative genre, one could expect a learning effect spilling from practices in doing talk for 'being descriptive' to doing talk for more involved, interpersonal purposes such as 'blaming', 'saving face', or 'seeking empathy'. This should not be misunderstood as meaning to imply that those latter purposes are learnt in more detached speech genres first. Not at all. But in cases where the linguistic procedures relevant for the construction processes of highly involved speech genres constitute a particular problem (such as in the case of constructing sad scenarios in the above), practices in more detached speech genres might enable speakers to sort out the procedures and re-integrate them at a higher level of integration in more involved speech genres.

    Taking up on a finding by Stein and her associates that was discussed above in section 1, namely that children younger than in our study were perfectly able to differentiate between the different components that distinguish (English) anger, sadness, fear, and happiness, which at the same time seems to contradict the findings reported in this section, we are now in a better position to reconsider this seeming contradiction and tie it closer to the concerns of methodology and development. While one of the important differences between the two studies under consideration is the age of the children, another one is the way the data were elicited. Apart from these two aspects, however, there is a third issue, which concerns the question of which aspects of performance we take to represent relevant developmental strides. Note that the on-line interview technique, used by Stein and her associates, is traditionally employed as a cognitive procedure to test comprehension. For this purpose it is legitimate to interrupt the natural conversational flow with questions that probe children's 'real' understanding. The discourse mode that is created in this type of interview resembles the way caregivers and children interact in a topic-elaborative style that is quite common in our culture, where the caregivers "build bridges" to test and teach 'knowledge'. This type of knowledge, though not necessarily of an abstract nature, is nevertheless accessed in a much more detached, quasi descriptive, explanatory mode. In contrast to this type of discourse mode, we found in our own investigation the early "confusion" of anger and sadness scenarios to be grounded in the involved discourse mode, where it was the primary goal to grammaticize the discursive force of the two different emotions. Thus, we do not see the findings of Stein and her associates contradicting ours. Rather, they complement our own findings in the sense that a more detached discourse mode in both investigations was found to facilitate a clear differentiation of what is underdifferentiated only in the involved first person genre.

    However, I assume that the relevance of these findings is weighted quite differently in the two different frameworks. A more cognitive oriented approach is most likely to consider as relevant the point in time when the 'knowledge base' for individual emotions (or other cognitive systems) first can successfully be tapped, because from thereon, all confusions can be (and need to be) explained in terms of situational performance constraints. In order to tap this 'basic' knowledge developmentally as early as possible, an elicitation technique must be found that imposes as little as possible situational and contextual constraints. Accordingly, within the cognitive framework of emotions, to ask children to construct emotion scenarios in the first person genre, might not count as an efficient way, exactly for the reason that their personal involvement might interfere with their actual knowledge. In addition, these "distortions" all occur after the basic knowledge of particular emotions has already been firmly established. Therefore, according to such a position, the above findings and our elaborate discussions do not contribute to how children establish their emotion knowledge. At best, they may contribute to how children "apply" their knowledge under difficult and adversary circumstances.

    In contrast to the cognitive framework of emotions and its approach to knowledge acquisition as the major developmental achievement, the discursive orientation views knowledge of the emotions rather as the result or the product of cultural learning and cultural practices than their starting point. The discursive approach to emotions is primarily interested in the processes through which cultural knowledge obtains its motivational force for individuals, and for this reason, a confusion between emotions in particular discursive settings is of utmost interest, since it offers insight into the developmental process of how the cultural directives of emotions are sorted out. Consequently, comparisons between performances in different practices or discourse settings is of extreme interest for studies which focus on development as a form of cultural learning. And the above findings and discussions present an important starting point within this orientation.

3. Emotions as linguistic construction types

One possibly puzzling concern lies hidden behind much of what has been discussed thus far: Why do people, when asked to give emotion accounts (of how they or others once felt), construe elaborate circumstances around happenings and events, i.e. seek refuge in the 'world of actions'? And why, when asked to construe events or happenings (in which they themselves or others figure as actors), do they deviate from the sequencing of actions and resort to references to feelings and emotions? This concern actually becomes more urgent when it comes to accounts in which the narrator seems to be implicated or implicating someone else, i.e. accounts that have been classified in the above as "involved". Edwards and Potter, in response to this seeming contradiction, argue that in natural discourse, talk about events and happenings is designed in particular ways "to allow inferences about mental life and cognition" (1992: 142) - and I would like to add, particularly inferences about emotions. Edwards and Potter also hold that the converse is equally true. Assumptions about the world, what happened and why it happened, are inferred from the way the speaker designed the emotions and motivations of the actors. Thus, event construal and the construal of characters' "inner psychologies" are closely orchestrated with regard to one another. Accounting for one, to a large degree, indexes the other.

    However, these worlds of actors and events within which references to emotions are embedded do not stand on their own. They are versions produced in discourse situations for discursive purposes. The particular purposes can be manifold such as to attribute blame to others (in order to save face or restore one's dignity), or to elicit empathy (in order to pursue joint retaliation or revenge). How the speaker "wants" 23 to be understood vis-a-vis the audience results in the particular construction design that is given to the world of actors and events. Thus, the world of the interactants (speaker-audience) functions as the regulator for the way the psychological "reality" of the actors are construed as well as what is happening in the events in which the actors have a role.

    As such, the approach I proposed in this paper attempts to turn around the traditional, realist picture of the relationship between emotions, cognitions, and language. While the realist picture starts from events and happenings as taking place in the world, to be re-presented in people's thoughts and feelings, so that we can subsequently speak about these events as well as the thoughts, evaluative appraisals, and feelings, thereby communicating them 24 , I am suggesting to invert this relationship: In communication, which is the performative domain of social action, both events as well as stances toward them (evaluative or cognitive) are organized; not because they are stored and available previous to and outside of the discursive purpose, to be executed subsequently in communication. Rather, events and the way they are thought about and valued by the person are constructs that are borne out of the purpose of talk. Consequently, the way in which the purpose of talk is manifesting itself in the world of interactants is not a by-product, but rather the starting point for a (discursive) analysis of what is manifested in the talk in terms of the characters, their activities, and the evaluative position with regard to them.

    This close relationship between the world of the characters and the world of interactants can equally well be described in terms of a positioning process that works both ways: By positioning the characters at the content plane with regard to one another, the speaker positions him/herself with regard to the listener; and this process works simultaneously the other way around. The coordination between these two planes results in the establishment of a moral position for which the speaker can be held accountable, irrespective of whether the speaker him/herself plays a role in what is being talked about, or whether the talk is merely about others.

    In sum, the approach to which I see this work contributing tries to systematically deconstruct two language functions that traditionally have been dealt with as independent domains: the referential/ideational and the communicative/interpersonal. While most semantic approaches rely heavily on the former, I see my own work as systematically extending Rom Harré's challenges to narrow, semantically based emotionologies. Inquiries into the discursive purposes of emotion terms necessarily have to take into account how a referential world is being construed in light of how the speaker-audience relationships are being construed. Thus, the content (of what is being constructed) and the purpose (for which content is being constructed) are closely linked by way of the linguistic construction types employed.

    Along with the deconstruction of the linguistic dichotomy between referential and communicative planes of language functioning runs a challenge to a deeply seated assumption of psychological theorizing: While it is traditionally assumed that the person has privileged access to his/her own psychological "inner states" (one's feelings, intentions, thoughts, and the like) (see Harris 1989, and Lubinski & Thompson 1993 for this position), the orientation adopted here runs counter to this assumption. Accounts that present emotion situations from the first-person perspective are analyzed in the same way as accounts that present a third-person or a generalized person perspective. In spite of the fact that the third-person perspective is often treated as the point of departure for fiction, there are no a priori grounds to give any of these perspectives a privileged status over the others. All of them are made out of linguistic constructions, the way illustrated above. And any difference in the way emotions (emotion talk) are constructed from these different person perspectives (first, third, generalized) is to be scrutinized in terms of what constructions are used for the processes of discursive positioning.

    As a concluding remark, let me briefly touch on an additional dimension regarding the notion of positioning that I alluded to in the above. In addition to the orchestration of characters with regard to one another at the level of what is being talked about (positioning level 1), and in addition to the orchestration of the speaker-audience relationship (positioning level 2), the same linguistic construction types index how the speaker positions him/herself with regard to the "Self" (positioning level 3). 25 Coordinating the content of talk with the discursive purposes for which it is constructed, forms the presupposition in what is commonly considered the construction of self-identity. Whether the talk is actually about the "self" (the I), or others (he or she), or about people in general (one or the generalized you), it always reveals aspects of a moral order in the way characters and audience are orchestrated (see Day & Tappan [1996] for linking this insight to the issue of moral development). The moral identity question "What am I in relation to the Good?" (Sarbin 1995, p. 219) turns into a position with regard to one's own identity: "Who am I?". Thus, the construction of characters in events at the level of content formation, the construction of the speaker-audience relationship, and the construction of one's self-identity are closely interwoven. And while traditionally psychologists start from the assumption of the unity of the "Self", and see narrative accounts and interactive relationships as orchestrated from this position, the approach that is schematically outlined in this paper views this unity of a "Self" (at least to a degree) as consisting of local achievements that are based on language use for discursive purposes (see Davies & Harré 1990 for a very similar orientation).

    In sum, what has been gathered together in this presentation is an attempt to more clearly delineate between cognitive approaches to emotions (and emotion development) and their discursive counterparts. What has come to the forefront in this attempt is the role of language, in particular of grammatical constructs (construction types) which form important building blocks in the formation of the social constructs that are achieved in development and interaction (eg. emotions, intentions, memory, self, and identity). The close look at some of these building blocks, as illustrated in the studies presented above, has led to an illumination of what is involved in the interplay of language, emotions, concepts of emotions, and in their respective developments. It is only hoped that this type of concern with linguistic forms does not achieve the opposite of what is intended, i.e. I hope it will not lead to an emotion that in turn might lead to a dismissal of the suggested approach as too linguistic, too technical, too complex, i.e. too scary.


References

    Bamberg, M. (1987). The Acquisition of Narratives. Berlin: Mouton de Gruyter.

    Bamberg, M. (1991). Narrative as perspective taking: The role of emotionals, negations, and voice in the construction of the story realm. Journal of Cognitive Psychotherapy 2, 275-290.

    Bamberg, M. (1993). Communication and internal states: What is their relationship? Behavioral and Brain Sciences, 16, 643-644.

    Bamberg, M. (1994a). Actions, events, scenes, plots and the drama. Language and the constitution of part- whole relations. Language Sciences, 16, 39-79.

    Bamberg, M. (1994b). Why narrativizing a wordless picture book is not artificial. - A response to Jim Gee's review of "Relating events in narrative". Teaching English as a Second Language. Electronic Journal, 1, 5-6.

    Bamberg, M. (in press a). A constructivist approach to narrative development. In M. Bamberg (ed.), Narrative Development - Six Approaches. Mahwah, NJ: Lawrence Erlbaum Associates.

    Bamberg, M. (in press b). Perspective and agency in the construal of narrative events. Proceedings of the 20th Boston University Conference on Language Development. Somerville, MA: Cascadilla Press.

    Bamberg, M. (in press c). Emotion Talk(s). The role of perspective in the construction of emotions. In S. Niemeier & R. Dirven (Eds.), The Language of Emotions. Amsterdam, The Netherlands: John Benjamins.

    Bamberg, M., Ammirati, N., & Shea, S. (1995). How children learn to talk about things with no name: Double emotions. In P. W. Davis (Ed.), Descriptive and Theoretical Modes in the Alternative Linguistics. Amsterdam: John Benjamins.

    Bamberg, M., & Damrad-Frye, R. (1991). On the ability to provide evaluative comments: Further explorations of children's narrative competence. Journal of Child Language, 18, 689-710.

    Bamberg, M. & Reilly, J. (in press). Emotion, narrative and affect: How children discover the relationship between what to say and how to say it. In D. I. Slobin, J. Gerhardt, A. Kyratzis, & J. Guo (Eds.), Social Interaction, Social Context and Language: Essays in Honor of Susan Ervin-Tripp. Mahwah, NJ: Lawrence Erlbaum Associates.

    Berman, R., & D. I. Slobin (Eds.). (1994). Relating Events in Narrative: A Crosslinguistic Developmental Study. Hillsdale, NJ: Lawrence Erlbaum Associates.

    Berman, R.A., Slobin, D.I., Bamberg, M., Dromi, E., Marchman, V., Neeman, Y., Renner, T. & Sebastian, E. (1986). Coding Manual: Temporality in Discourse (Revised Edition). Institute of Cognitive Studies, University of California, Berkeley.

    Besnier, N. (1994). Involvement in linguistic practice: An ethnographic appraisal. Journal of Pragmatics, 22: 279-299.

    Budwig, N. (1995). A Developmental - Functionalist Approach to Child Language. Mahwah, NJ: Lawrence Erlbaum Associates.

    Buttny, R. (1993). Social Accountability in Communication. Newbury Park, CA: Sage Publications.

    Caffi, C., & Janney, R. W. (Eds.) (1994). Involvement in language. Special issue of Journal of Pragmatics, 22.

    Capps, L. & Ochs, E. (1995a). Constructing Panic. The Discourse of Agoraphobia. Cambridge, MA: Harvard University Press.

    Capps, L. & Ochs, E. (1995b). Out of place: Narrative insights into agoraphobia. Discourse Processes, 19, 407-439.

    Campos, J. J., Mumme, D. L., Kermoian, R., & Campos, R. G. (1994). A functionalist perspective on the nature of emotion. In N. A. Fox (Ed.), The Development of Emotion Regulation: Biological and Behavioral Considerations. Monographs of the Society for Research in Child Development, 59, (Nos 2-3), pp. 284-303.

    Davies, B., & Harré, R. (1990). Positioning: The discursive production of selves. Journal for the Theory of Social Behaviour 20, 43-63.

    Davitz, J. (1969). The Language of Emotions. London, England: Academic Press.

    Day, J.M., & Tappan, M.B. (1996). The narrative approach to moral development: From the epistemic subject to dialogical selves. Human Development, 39, 67-82.

    Edwards, D., & Potter, J. (1992). Discursive Psychology. London, England: Sage Publications.

    Ellsworth, P. C. (1994). William James and emotion: Is a century of fame worth a century of misunderstanding? Psychological Review, 101, 222-229.

    Fiehler, R. (1990). Kommunikation und Emotion. Berlin: De Gruyter.

    Fischer, A.H. (1991). Emotion Scripts. A Study of the Social and Cognitive Facets of Emotions. Leiden, The Netherlands: DSWO Press.

    Geertz, C. (1973). The Interpretation of Culture. New York: Basic Books.

    Goddard, C. (in press). Reacting to the unexpected: Contrastive semantics of Malay and English. Culture and Psychology.    

    Harré, R. (1986). An outline of the social constructionist viewpoint. In R. Harré (Ed.), The Social Construction of Emotions. Oxford, England: Basil Blackwell.

    Harré, R. (1993). Universals yet again: A test of the 'Wierzbicka thesis'. Language Sciences, 3, 231-238.    

    Harré, R., & Gillett, G. (1994). The Discursive Mind. Thousand Oaks, CA: Sage Publications.

    Harris, P. L. (1989). Children and Emotion. Oxford: Basil Blackwell.

    Heelas, P. (1986). Emotion talk across cultures. In R. Harré (Ed.), The Social Construction of Emotions. Oxford, England: Basil Blackwell.

    Holland, D. & Quinn, N. (Eds.) (1987). Cultural Models in Language & Thought. New York: Cambridge University Press.

    Hopper, P. & Thompson, S. (1980). Transitivity in grammar and discourse. Language, 56, 251-299.

    Labov, W. & Fanshel. D. (1977). Therapeutic Discourse. New York: Academic Press.

    Leff, J. (1973). Culture and the differentiation of emotional states. British Journal of Psychiatry, 123, 299- 306.

    Lubinski, D., & Thompson, T. (1993). Species and individual differences in communication based on private states. Behavioral and Brain Sciences, 16, 627-680.

    Lutz, C.A. (1982). The domain of emotion words in Ifaluk. American Ethnologist, 9, 113-128.

    Lutz, C.A. (1988). Unnatural Emotions: Everyday Sentiments on a Micronesian Atoll and their Challenge to Western Theory. Chicago: Chicago University Press.

    Lutz, C. & White, G.M. (1986). The anthropology of emotions. Annual Review of Anthropology, 15, 405- 436.

    Matsumoto, D. (1994). People. Psychology from a Cultural Perspective. Pacific Grove, CA: Brooks/Cole Publishing.

    Moffet, J., & McElheny, K. R. (Eds.). (1966). Points of View: An Anthology of Short Stories. New York: New American Library

    Much, N.C. (1992). The analysis of discourse as methodology for a semiotic psychology. American Behavioral Scientist, 36, 52-72.

    Oatley, K. (1991). Living together. A review of 'Unnatural Emotions', by Catherine A. Lutz. Cognition and Emotion, 5, 65-79.

    Sarbin, T. R. (1995). Emotional life, rhetoric, and roles. Journal of Narrative and Life History, 5, 213-220.

    Shotter, J. (1993). Conversational Realities. London, England: Sage Publications.

    Shweder, R.A. (1991). Thinking through Cultures: expeditions in Cultural Psychology. Cambridge, MA: Harvard University Press.

    Stearns, P. N., & Stearns, C. Z. (1987). Emotionology: Clarifying the study of the history of emotional standards. American Historical Review, 90, 813-830.

    Stearns, C. Z., & Stearns, P. N. (1988). Emotion and Social Change. New York: Holmes and Meier.

    Stearns, P.N. (1994). American Cool: Developing a Twentieth-Century Emotional Style. New York: University Press.

    Stearns, P.N. (1995). Emotion. In R. Harré & P. Stearns (Eds.), Discursive Psychology in Practice. London, England: Sage Publications

    Stein, N.L. (1979). The concept of a story: A developmental psycholinguistic analysis. American Educational Research Association, San Francisco, CA.

    Stein, N.L. (1988). The development of storytelling skill. In M. Franklin & S. Barten (Eds.), Child Language: A Reader. New York: Oxford University Press.

    Stein, N.L. & Albro, E.R. (in press). Building complexity and coherence: Children's use of goal-structured knowledge in telling good stories. In M. Bamberg (Ed.), Narrative Development - Six Approaches. Mahwah, NJ: Lawrence Erlbaum Associates.

    Stein, N.L. & Trabasso, T. (1992). The organization of emotional experience: Creating links among emotion, thinking and intentional action. In N. Stein & K. Oatley (Eds.), Cognition and Emotion (special issue), 6, 225-244.

    Stein, N.L., Trabasso, T. & Liwag, M. (1994). The Rashomon phenomenon: The role of framing and future- orientation in memories for emotional events. In M.M. Haith, J.B. Benson, R.J. Roberts, Jr., & B.F. Pennington (Eds.), The Development of future oriented processes. Hillsdale, NJ: Lawrence Erlbaum Associates.

    Stein, N.L., Liwag, M.D., & Wade, E. (in press). 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.), Emotions: The G. Stanley Hall Symposium. Hillsdale, NJ: Lawrence Elbaum Associates.

    Talbot, J., & Bamberg, M. (in press). Affirmation and resistance vis-à-vis the predominant discourse: The rhetoric construction of pregnancy. Journal of Narrative and Life History, in press.

    Trabasso, T. & Stein, N.L. (1993). How do we represent both emotional experience and meaning? A review of Richard S. Lazarus' "Emotion and Adaptation". Psychological Inquiry, 4, 326-333

    Trabasso, T., Stein, N.L., Rodkin, P.C., Munger, M.P., & Baughn, C.R. (1992). Knowledge of goals and plans in the on-line narration of events. Cognitive Development, 7, 133-170.

    Walters, K. (1995). Review of Anna Wierzbicka: 'Cross-cultural semantics: The semantics of human interaction'. Language in Society 24, 564-569.

    White, G. M. (1990). Moral discourse and the rhetoric of emotions. In C. A. Lutz, & L. Abu-Lughod (Eds), Language and the Politics of Emotion. Cambridge, England: Cambridge University Press, pp. 46-68.

    Wierzbicka, A. (1992). Defining emotion concepts. Cognitive Science, 16, 539-581.

    Wierzbicka, A. (1994). Emotion, language, and cultural scripts. In S. Kitayama & H. R. Markus (Eds.), Emotion and Culture. Washington, DC: American Psychological Association, pp. 133-196.

    Wierzbicka, A. (1995a). Emotion and facial expression: A semantic perspective. Culture and Psychology, 1:227-258.

    Wierzbicka, A. (1995b). Dictionaries vs. encyclopedias: How to draw the line. In P.W. Davis (Ed.), Alternative Linguistics: Descriptive and Theoretical Models. Amsterdam, The Netherlands: John Benjamins.

    Wierzbicka, A. (in press). Response to Michael Bamberg. In S. Niemeier & R. Dirven (Eds.), The Language of Emotions. Amsterdam, The Netherlands: John Benjamins.

    Wittgenstein, L. (1953). Philosophical Investigations. New York: Macmillan

    Young, K. (1987). Tale Worlds and Story Realms. Dordrecht, Holland: Nijhoff.


TABLE 1:    TRANSITIVITY RATINGS

        ["Actor Transitivity" in the First Person Genre]

             HIGH Transititvity             LOW Transitivity

        BEING ANGRY        BEING SAD     BEING ANGRY    BEING SAD

Preschool     84%         65%            6%         0%

Kindergarten     70%         60%            7%         15%

1st Grade     86%         48%            6%         9%

3rd Grade     80%         32%            0%         38%

TABLE 2:    BEHAVIORAL SYMPTOMS

             Being ANGRY             Being SAD

         1ST PERSON EXPLANATION     1ST PERSON     EXPLANATION

Preschool        0%         2%            10%         35%

Kindergarten        0%         0%             0%         10%

1st Grade        0%         0%            25%         20%

3rd Grade        0%         0%            35%         50%

TABLE 3:    EXPLICITNESS OF UNINTENTIONALITY

             First Person Genre             Explanation Genre

         Being ANGRY        Being SAD     Being ANGRY     Being SAD

Preschool        6%         20%             0%            50%

Kindergarten        8%         20%             0%            42%

1st Grade        6%         23%             0%            25%

3rd Grade        10%         50%             0%            67%

TABLE 4:    EXPLICITLY UNJUST ACTIONS

              First Person Genre                 Explanation Genre

         BEING ANGRY     BEING SAD             BEING ANGRY     BEING SAD

Preschool        29%         0%                30%             0%

Kindergarten        23%         7%                 7%             0%

1st Grade        29%         8%                 0%             0%

3rd Grade        53%        13%                17%             0%

TABLE 5:    TRANSITIVITY RATINGS

        ["Overall Transitivity" in First Person and Expanatory Genre]

             HIGH TRANSITIVITY            LOW TRANSITIVITY

                        BEING SAD

         1ST PERSON     EXPLANATION     1ST PERSON EXPLANATION

Preschool        70%         44%               30%        55%

Kindergarten        80%         40%              20%        60%

1st Grade        65%         43%              35%        57%

3rd Grade        37%         14%              63%        86%

Converted by Andrew Scriven