CHAPTER 4

When psychologists, the non-linguists who are specialists in emotion, analyse the language of Emotion, they nearly always focus the lexicon. As we shall see, the lexicon of Emotion cannot be defined rigidly, and one has to allow for a certain fuzziness in distinctions between what is, or is not, an Emotion, as well as degrees of classification, for example between basic and peripheral emotions, particularly when the context in which the lexeme occurs is taken into account.

It is easy to see that some will feel that the Platonist tradition and the idea of innate concepts is relevant to the analysis of Emotion. Since emotions are seen as basic to human behaviour, with roots in earlier stages of evolution, it is understandable that they should be candidates for the status of innate concepts. Our expression of our own emotions and our perception of emotions in others is possibly more frequently communicated by behaviour rather than language, and this behaviour is arguably pre-lingual both phylogenetically and ontogenetically. For example, the emotions which cause one to smile when one is happy, or to scream with fear, seem to be universal to the human species and to develop in individuals in a fairly predictable way. The ways they are expressed have recognisable counterparts in animal behaviour. Emotions, therefore, would seem to be fairly primary concepts coming perhaps somewhere between the physically generated Hunger, Thirst, and Sexual attraction, which can be seen as essential for the survival and propagation of even the humblest species, and those concepts so many people would like to consider universal, Goodness, Beauty and Truth.

When one is talking about emotion in one's own language, it is very easy to feel that such concepts must exist universally. Everyone, one feels, must know what it is like to be happy or sad, or have felt love or hate - after all, when watching someone on television, whose language we do not understand, protesting about some injustice, we may need someone to inform us about the injustice, but we can deduce the people's emotions about it from their facial expressions and gestures. Or can we? Maybe we can deduce some general negative notion about the injustice, but we may need to understand what is being said before we can be clear whether the dominant emotion being expressed is anger, hate or frustration, and it is quite possible that the relevant emotion words in the unknown language may present problems for whoever is translating or interpreting for us. The language relativists will find that the lexicon of Emotion can be shown to vary between different cultures, languages and social groups. As we shall see, psychologists and linguists have in fact found that individuals using the same language vary in their usage of Emotion lexemes more significantly than in other lexical areas.

For centuries, attempts to categorise emotion have been made by all those interested in the subject. As we have already seen, Aristotle, Spinoza and Descartes expounded at length on the subject, and others such as Locke and Hume also contributed to the debate. Since James and Darwin, several different classifications have been tried, some reducing the categories to two basic ones such as negative and positive emotion, or pleasure and pain, and others elaborating long lists. Davitz (1969), a psychologist, compiled a 'dictionary' of 50 terms. Wierzbicka (1972), a lexicologist, attempted 36 definitions of basic emotions based on situations, although by (1992: 45) she is proposing a more sophisticated approach. Another, more linguistic, approach is that of those like Lakoff and Johnson (1980), who propose that we can arrive at conclusions about the conceptual framework of experience by looking at the metaphors we use to describe this experience, rather than relying on the the more traditionally accepted terms. Johnson-Laird and Oatley (1989), psychologists/AI experts, have made a list of several hundred of the accepted emotion words, which they subdivide into 5 basic emotion modes and 6 semantic categories. Ortony, Clore and Collins (1988), psychologists with an interest in AI and language, propose a fairly complex categorisation based on what they consider to be the cognitive structure of emotions, rather than actual words. The last three groups all describe their approach as cognitive.

The reason for defining a set of basic emotions is partly due to a need for economy in an area which is prolific in lexemes. Even those who draw up such shortlists sometimes vary in their choice of synonym, preferring perhaps anger to rage. This is bound to happen, not least because some of these choices are governed by influences from other languages. However, the choice is also affected by the theory of emotion on which it is based.

Kenny (1963) describes Descartes' shortlist of Gladness and Sadness, Love and Hatred, and Desire, as being designed on the basis that the other emotions were built out of combinations of these - Pride was a compound of Gladness and Love, and Pity one of Sadness and Love - and he worked out a complex physiological explanation for each emotion. This approach is echoed in that of modern psychologists. James (1884) chose 4 strong emotions, Fear, Grief, Love and Rage, which were selected on the basis that they needed a "distinct bodily expression". Other theorists also formed lists based on physiological change, both those analysable by internal, visceral change and by external expression. Ekman, Friesen and Ellsworth (1972) drew up a list based on a cross-cultural study of facial expression which gave them Anger, Disgust, Fear, Joy, Sadness and Surprise.

Arnold (1960) was one of the first to propose a selection on a more cognitive basis, as she saw the emotions as based on a relation to the action tendencies for which the perception of emotion prepared the body. Fridja (1986) also chooses his list, Desire, Joy, Pride, Surprise, Distress, Anger, Aversion, Contempt, Fear and Shame, on the basis of forms of action readiness. Other cognitivists have attempted classifications based on a view that certain emotions are 'hardwired' into the neurological system. Tomkins (1962; 1963) has taken the view that affect is based on motivation, Plutchnik (1962 -1982) developed his list based on a psycho-evolutionary viewpoint, and yet others have based their views on the observation of the development of emotions in infants, children and adults. Others, like Izard (1991), have taken several of these factors into consideration to produce lists which they believed to be either hard-wired into the nervous system, or partly so, in a way that is later influenced by developmental and cultural components. Johnson-Laird and Oatley (1989) chose their list of basic emotions, Happiness, Sadness, Fear, Anger and Disgust, using tests of a more linguistic nature, as we shall see, but Ortony, Clore and Collins (1988) prefer to classify emotions according to certain criteria which do not include the notion of a basic emotion.

Lexicons of emotion are to be found in thesauri and general lexicons. Roget' Thesaurus (1962 edition), for example, lists them under Class VI - 'Affections', and sub-divides this general class into general, personal and sympathetic affections. However, it is the psychologists and AI experts who seem to have dedicated most effort to disentangling this complex lexical field. Therefore, let us look at some of the more linguistically orientated attempts at categorising emotion and related lexical areas.

Davitz (1969) set out to produce a dictionary of emotion by collecting lexemes as they are used in a non-technical, everyday fashion. He first selected about 400 words from Roget's Thesaurus and then asked 40 subjects to give their opinion as to which of these words constituted an emotion. Of the 137 chosen, he then selected the most representative 50, and made a check list of 556 possible descriptive statements of the physiological and psychological feelings that are associated with emotions. 50 subjects were then asked to select the statements which they felt most suitably described each emotion.

After analysing the results, 215 of the descriptions, which had appeared significant enough to appear in the analysis of at least 3 emotions, were classified into 12 clusters which expressed certain physical sensations or reactions associated with Emotion : activation, hypoactivation, hyperactivation, moving toward, moving away, moving against, comfort, discomfort, tension, enhancement, incompetence / dissatisfaction, and inadequacy. The emotion terms were then analysed as to the relative effect of each of these clusters on their definition. Thus, for example, a proportion of the definers classified anger as related to hyperactivation (53%), moving against (46%), tension (19%) and inadequacy (16,8%); depression as related to hypoactivation (48,6%), moving away (38%), discomfort (41,8%), tension (5,4%), incompetence / dissatisfaction (19%) and inadequacy (35,2%); and love as related to activation (29,6%), hyperactivation (12%), moving toward (66%), comfort (32,2%), enhancement (17,6%) and inadequacy (6,7%) (ibid : 116-119)[1].

The results of this experiment were varied. 11 of the 50 terms "did not achieve adequacy ratings indicative of comprehensive and accurate definitions" (ibid : 86), which means that that those questioned described the terms in such different terms, that the researcher was unable to produce a significant consensus of opinion on them from his data. This shows up particularly in the cluster analysis where dislike, for example, is seen as related to moving against (14,8%), tension (10,4%) and incompetence/dissatisfaction (5%), hardly a satisfactory result. In the 'dictionary' it is noticeable that agreement over the descriptions used varies considerably, usually as low as 34% but descending to 10%, and reaching a maximum agreement of 84%, although only a small percentage achieve over 50% agreement. These factors may have been due to the differences of opinion among the experimentees, but Davitz felt it more likely to be some inadequacy of the check list. This is possibly true, because the checklist emphasises physiological feelings, and several of the 11 terms require a certain measure of more conscious cognitive processes - gratitude and resentment, for example.

One must remember that Davitz was working in the early days of cognitive psychology when physiological reactions were still considered the only scientifically reliable ones. He describes how aspects of his theory have been tested on different social groups - a developmental study of the language of emotion in children, individual differences in emotional reports as a function of perceptual-cognitive styles in adults, similarities of emotional descriptions in relation to genetic background (i.e. by studying twins), and a cross-cultural comparison of the language of emotion among adolescents in the United State and Uganda.

The list of 50 emotional states in itself is representative and the very data that emerges could be used as an argument for reducing the number to a more basic set of concepts. For example, cheerfulness, delight, gaiety and happiness, show a fairly regular correlation as regards cluster scores and could be seen as variations on a common theme. Certain terms could be questioned. Passion has been included, although some would see it as a generic term, like emotion, or at least as referring to stronger emotions like fury and love. Interestingly, Davitz's experimentees, who were American, seem to identify it largely with sexual love. Others, like friendliness, impatience and solemnity, appear more frequently in the context of behaviour resulting from emotion than in that of a definite emotion.

The 12 clusters can be seen to form 4 dimensions of emotion: activation which other psychologists and phenomenologists had described previously as 'level of activation', 'emotional energy', or had mentioned as excitement in contrast with quiescence and depression; relatedness, which emphasises the Subject's relation with the outside world, sometimes described as direction by others; hedonic tone, seen by others as the pleasant-unpleasant factor; and competence, which fits more closely, either with the view that emotions evolve in response to an evolution of the organism's needs in the environment, or with that of those, like Sartre, who see emotion as the individual's way of adapting to reality, or trying to change the reality to allow for this adaptation.

At the time, Davitz's work was considered as lacking in empirical data and rather too phenomenological for the taste of psychologists, while lacking the formalisation required by AI. However, in the cognitive climate, twenty years later, more importance seems to have been given to his work. Fridja (1986 : 184) apologises for his negative review of the book when it was published, and goes on to describe it as "an original and valuable study". Strongman, whose textbook The Psychology of Emotion has been substantially revised three times as his early Behaviourist stance changed to a more Cognitivist one, is decidedly more positive about Davitz's work in the 1987 edition of his book than he had been in 1978. Davitz himself only hoped that "this report represents a minor beginning in a direction of research that could conceivably have some implications for a broad range of psychological and social problems". As other studies bear out, this is not an easy lexicon to study and, although there is enough similarity between the opinions of the subjects tested on which to build a theory of emotion, the differences of opinion between individuals are wide enough to make these differences of interest.

This study forms part of a wider project in Structures in the Subjective Lexicon (1971) and the data was collected using criteria from both psychology and linguistics. The psychologists who have influenced this study most are those, like Ekman, who use facial expression as their point of departure. The linguistic approach owes most to structural semantics, particularly as regards semantic and lexical fields. Using 105 pairs of emotion terms, they asked 60 subjects, divided into 4 groups, to "rank pairs of emotion names with regard to a similarity or dissimilarity criterion, to build a similarity tree using the names....., as well as requiring subjects to rate directly the similarity between every pair of terms". (ibid : 110)

The data thus obtained was examined using graph analysis, cluster analysis, and multidimensional scaling analysis, methods they had applied to other areas of the lexicon. They found that the difference between individuals and between the four control groups were larger for the lexicon of Emotion than in the other semantic areas they had studied. Furthermore, "attempts to interpret individual representations or to classify subjects according to their representations were, by and large, fruitless"(ibid : 122). The three types of analysis yielded rather similar results, so it was possible to argue that there was little wrong with the methods used to produce some sort of order, but the difficulty was to decide what lay behind this order. They suggest that, once people get beyond the "very gross contrasts such as that of Pleasantness-Unpleasantness", discrimination between emotion names becomes rather idiosyncratic (ibid: 123).

The authors agree that their analysis of emotion did not lead to any substantive conclusions. They feel that:

They conclude that their work "might be regarded as constituting fairly impressive evidence against the adequacy of the various sorts of theoretical conceptualizations that have been offered for the realm of emotion or emotion names", but say that the problems posed by individual differences "suggest[s] that what may be required is not a theory for the domain of emotions, but perhaps a set of characterizations whose members may differ systematically in important ways" (ibid : 124).

These attempts to get people to describe the (usually) physical feelings they experience in emotional situations run parallel to efforts by certain linguists, who propose to analyse the meaning of emotion using semantic components of the type that can be related to the more physically verifiable cognitive processes. This approach owes much to Lakoff and Johnson (1980), and is demonstrated more specifically in relation to Anger in Lakoff and Kövecses (1987). This type of linguistic investigation is similar to that carried out by Davitz, the difference being that whereas Davitz was trying to see how people linked the description of certain physical and psychological feelings to certain Emotion concepts for reasons linked to psychology, the linguists are using similar materials to prove something about language.

According to this type of theory, as Barbara Lewandowska (personal communication) explains, one may take a notion such as 'down', which has prepositional and directional qualities which can actually be described in non-linguistic terms with a diagram, and show how depressed, and other emotion types of what others might describe as Distress can be described in terms of 'down-ness'. It would seem that there is an association in several languages of these emotions with the literal notion of down-ness which has probably developed from the body language with which we physically express these emotions - we look down, the corners of our mouths turn down, our bodies seem to move earth-wards, and there is a general impression of almost physical pressure from above which we produce in metaphors like he was weighed down by his problems. Conversely, positive emotions are associated with up-ness because our expressions lift, we stand erect, and we look up on these occasions.

Alternatively, it will take an expression like boiling with anger and, rather than simply describe the boiling as a metaphor for anger, it will show how the abstract anger can only be explained by associating the physical sensation of heat and agitation we feel as our blood pressure rises in these circumstances with the more easily observable phenomenon we observe when our kettle prepares water to make tea. I have also seen work of this type in a paper by Fernando (1989). For example, it analyses the way we describe our emotions, listing expressions such as be in a cold sweat and tremble under Fear and see red and blow one's top under Anger (p.11). The author also makes observations such as "the opinion implicit in Jack is evil may simply be another of saying I dislike Jack" (p.2), an analysis which is based on more pragmatic and situational criteria than the type of analysis I propose to make here.

This type of cognitive approach tends to use those metaphors which refer to physically identifiable phenomena in the belief that, by so doing, it will facilitate the eventual codification of abstract concepts. If one believes that the type of input that the brain receives through physical perception stands a chance of being codified in a way artificial intelligence can understand in the not-too-distant future, perhaps this approach will eventually go beyond the more mechanical computer database and help the 'neural network' type of computer understand how human beings function, presumably at a vicarious level. Artificial intelligence will only ever experience this type of physical perception when biological science manages to come up with the type of physically human robot + artificial brain imagined in the final chapter of Asimov's "I, Robot".

In the meantime, even if this meaning is deduced from original texts, or appropriate tests on individual speakers describing their personal experiences of the emotions and consequent understanding of the words, the tendency is necessarily reductionist because it favours analysis of the physical perception input which precedes appraisal. As a linguistic theory it follows the trend of certain psychologists to continue to assess psychological phenomena from observable physical symptoms. These psychologists often call themselves 'cognitivists', because they no longer rely on mechanical stimulus-response experiments and allow for the more complex intervention of the subject's own interpretation of events, but they still use a lot of pre-suppositions they have inherited from their Behaviourist predecessors.

It should also be noted, as Lutz (1987: 292) points out, that to describe emotion concepts in terms of "physical and / or private mental states" may be fairly typical of American English, but is not the norm in other cultures. She describes, for example, how the Ifaluk islanders prefer to describe the social or cultural situation that leads to the identification of the emotion, and rarely refer to their own personal reactions. This obviously points to one of the underlying themes of her research - that American English reflects the Self-orientated culture it is spoken in, whereas Ifaluk is spoken by a closely -knit communal society,

A less mechanistic cognitivist, even if s/he is a speaker of American English, would try to account not only for the more immediate physical perceptual input, but also for all the other information stored in the brain which influences the appraisal of this input, including the linguistic codification assigned to such input by previous experiences, both personal and vicarious. It is possible that, with the help of such codification, the brain perceives the input not so much as a series of components based on physical perception, but rather as a type of more holistically viewed experience, or 'scenario'. This is the type of analysis which some of the other writers discussed in this chapter prefer.

Wierzbicka's (1972) approach is that of a lexicologist writing about Semantic Primitives and how they may or may not provide help in defining terms. The chapter in this book devoted to the emotions starts with the quotation from Wittgenstein's Philosophical Investigations - "How is one to define a feeling? It is something special and indefinable". She herself goes on to say that "an emotion is something that is felt and not conceived verbally" and that "thoughts have a structure which can be rendered in words, but feelings, like sensations, do not" (ibid : 59). The best we can do is to try and describe what we feel, and depend on other people's experience of similar things to help them understand us. She quotes in support of this several examples from Tolstoy's "Anna Karenina" when he compares what X felt to something else, using phrases such as felt + as / rather as / like.

She suggests that this is a useful technique with the more complex emotions, but goes on to question whether Emotion words such as joy, sadness, anger and fear can be considered as semantically simple, just because they are more common. After all, if they were that simple, they would not even be related to each other, as they obviously are. If they are not simple, or primitives, it is therefore possible to define them, although perhaps only by reference to the situations associated with them[2].

She suggests that even the commonest emotion words are really "petrified, disguised abbreviations of descriptive formulae", and goes on to make 36 definitions of some of the more common emotional states. She uses the expression "X feels", and the reference is more frequently to the adjective than the noun form of the emotion lexeme. This facilitates the connection feeling <--> emotion, but causes a certain irregularity in her examples. For example, she uses envious, but not envy, and compassion, but not compassionate, and it has

She compares synonyms like want and desire, and fear and anxiety, in order to establish that the first of each pair is somehow more basic than the other. She also draws comparisons based on the symmetrical relations between examples like satisfied and dissatisfied, and shows that a similar situation exists with joyful and upset. These comparisons are not always so straightforward, though. Pity can be related in this way to envy, contempt and admiration, depending on the component involved. She argues, too, for components like [GOOD] and [BAD] and [INTRAPERSONAL] to explain the relations between other lexemes.

This is a short essay written rather to arouse interest in a certain lexical field than as a detailed study designed to solve any fundamental problem. I quote from it rather because of the influence it had on the debate rather than as a reflection on her latest ideas. In Wierzbicka (1992) her approach is more sophisticated, and she has become involved in the discussion at the level of psychology and the theories put forward by the next two theorists to be considered. Here she acknowledges that the search for semantic primitives must be more rigorous than was possible in 1972. Since then, she claims, work has been done to provide a "Natural Semantic Metalanguage" which "has been tested in hundreds of definitions, both of English words and of words from many other languages of the world. It has also been tested in the study of grammar and cross-cultural pragmatics. She believes that "what is needed is a body of definitions large enough to be able to reveal the systematic organization of this cognitive domain and to demonstrate conclusively that the proposed method of definition really works". Rather than work out these definitions in terms of individual lexemes, beyond the bare minimum needed, she argues in favour of "certain prototypical scripts or scenarios, formulated in terms of thoughts, wants and feelings"[3]. When analysing the controversies in the more recent literature on the emotions she suggests (1992:17) that:

This is a particularly interesting point, especially if we think of the individual differences of usage suggested by the research of Davitz and Fillenbaum and Rapoport. Wierzbicka herself, while defending the idea of some form of universal semantic primitives, also believes that some kind of "social construction of emotions" ... is largely right, and that different languages provide concepts which in their turn "provide certain 'scripts' which native speakers can use as a basis for interpretation of feelings and upon which they can model their emotions and their relations to others." Being a bilingual, or multilingual, herself though, she feels she cannot insist enough that although people make proposals about Emotion in their own language, these proposals are only valid insofar as they show awareness of the limitations posed by any language and allow for the fact that other languages see the same apparent reality through a different prism. She also points to the fact that "most authors writing about emotion concepts are content to mention just a few examples, in an ad hoc fashion" although she cites the next two groups of theorists as "two notable recent exceptions" and she admits a need for the type of analysis they make. Even so, she has a criticism to make of their definitions which, by her standards, are decidedly sloppy. Being psychologists, they cannot be expected to be aware of the full intricacies of linguistic, and particularly semantic. theory, any more than a linguist can be fully informed on the finer points of psychology. Probably all involved would agree that each can benefit from the other's help.

These are all points on which I am in full agreement with her, and, if my purpose were to set up universal hypotheses for semantic primitives of Emotion, I might turn to her for advice, despite my scepticism about the possibility of creating a Natural Semantic Metalanguage with which to do this. Although I see the attractions of it theoretically, I cannot believe that even the most objective attempts at such a metalanguage can go very far without being enmeshed in its own reflexivity. However, my interests are related rather to analysing what I find in the context of normal language use and at levels other than that of the lexeme.

4.3.5 Johnson-Laird and Oatley (1989) : a Cognitive lexicon which assumes basic emotions

Towards the end of the 80s, the interest in the language of emotion intensified, as does the psychologists' interest in language generally, thanks in large part to the direction being taken by Cognitive Science. Johnson-Laird and Oatley belong to one group of investigators into the language of emotion, and Ortony, Clore, Collins and others to another, and their work, and that of others, are the subject of several publications. They have a certain amount in common, and their differences tend to serve the cause of even more careful investigation, rather than destructive criticism .

Using language that shows their AI interests, Johnson-Laird and Oatley propose that "the human cognitive system is modular and asynchronous", and that:

They postulate five emotion modes, or basic emotions, that correspond to Happiness, Sadness, Anxiety (or Fear), Anger and Disgust. They believe that these modes serve as a basis from which more complex emotions are elaborated. Since it their belief that emotions play an important role in human communication and planning, they believe that "the cognitive system adopts an emotion mode at a significant juncture of a plan" and that "the function of these modes is to organise a transition to a new phase of planned activity directed to the priorities of the mode with associated goals and certain stored plans for dealing with what has happened" (ibid : 35) The way the system works is partly due to evolutionary factors, but also dependent on how the individual, as he/she develops according to both innate and environmental factors, forms a model of Self and Other, so that, with adult emotion, a social plan is a typical part of the way we experience emotion. Complex emotions "are founded on a basic non-propositional emotion mode, but have propositional evaluation which is social and includes reference to the model of self" (ibid : 46).

In Johnson-Laird and Oatley (1989), the focus is on the more linguistic criteria they used to compile the lexicon. They collected their 590 words from various sources and, while recognising that most emotion lexemes have several forms, they seem to choose the form which, to them, is most related to emotion, only referring to another form if it tends to have a different connotation. In an attempt to establish some formulae for semantic primitives, they suggest that the respective concept can be chosen on the basis that the emotion involved can be experienced without the experiencer knowing its cause :I feel x but I don't know why. Their system of establishing semantic relations between the basic emotions and their synonyms, and distinguishing between different emotions, is to use linguistic tests like those used by Bendix (1966). They also suggest that the temporal duration of emotions should be taken into account to allow for a distinction between emotions, moods and personality types. They distinguish between 'mood' and 'emotion', the former being of longer duration and in which the Object is less easily definable, if at all, than in the case of an emotion, which they consider to be typically short-lived. They believe that "the vocabulary of basic emotions should provide descriptions of moods and personalities".

Having decided on the five Basic emotions, they classify the more complex emotions into the five further categories:

There is also a zero category for Generic emotions, such as emotion and feeling.

Ortony and Clore (1989), while agreeing over certain general principles, criticise both the idea of basic emotions, which they reject in their own analysis. and the criteria used to distinguish them from other emotions. They do not like the I feel x but I don't know why formula, partly because the hypothesis is weakened if one substitutes feel by be, and partly because they believe that some of the examples given are questionable. Oatley and Johnson-Laird (1990), counter with the argument that Ortony et al.'s theory of emotions as valenced reactions "appears to embrace too much besides emotion", and that perhaps both parties are guilty of allowing their own intuitions govern their notions of what is and is not acceptable. However, they concede that perhaps the inclusion of a sixth emotion mode, Desire, as a pair for Disgust, and as the mode that would designate the otherwise isolated lexical category of Emotional goals, would improve their analysis, although the disadvantage of doing this is that "Desire is normally experienced in relation to an object, thereby implying propositional content", and undermining the I feel x but I don't know why theory.

4.3.6 Ortony, Clore and Collins (1988) : a Cognitive analysis of the emotions as valenced reactions

Ortony et al. started started their research in about 1981, shortly after Donald Norman (1981) identified the topic of emotion as one of the twelve major challenges to cognitive science. Encouraged by the fact that cognitive psychology had shown that emotion entailed appraisal through cognitive processes, they prepared the research that is the subject of their work, The Cognitive Structure of the Emotions. The two goals they hope to achieve are, first, "to bring some semblance of order to what remains a very confused and confusing field of study", and second, to try to "lay the foundation for a computationally tractable model of emotion". (Ortony et al. (1988 : 2))

This approach assumes some sort of Platonic reality for the emotions, while recognizing the language 'trap'. This, they claim, makes their account, "in principle, capable of accommodating the fact that there are significant individual and cultural differences in the experience of emotions". In this respect they have learnt from reading Wierzbicka and others. As a result they are careful to define the limits of the theory and not to claim that they are offering a total view of the subject.

While recognising the importance of behaviour and physiology to theories of emotion, they propose to pay minimal attention to these aspects in order to concentrate on the cognitive aspects which, they believe, precede the behavioural and physiological aspects. Cognition, in their view, may or may not imply consciousness of the cognitive processes, for, as Freud believed, the beliefs or cognitions on which emotions are based can be unconscious, ........, but the emotions themselves cannot be unconscious". (ibid : 176) It is also true that some emotions, like Disgust, require less cognitive processing than others, like Shame. Certain languages will have a greater capacity for expressing certain emotional areas than others, but this does not mean that that the other areas, which may be minimised or even ignored in a particular language, should not be taken into account. For this reason, they propose to establish emotion 'types', and then use 'tokens' from, in their case, English to exemplify these types.

Their attitude to 'self reports' of emotion is similar to that of Descartes - "that if a person..... is experiencing fear, that person cannot be mistaken about the fact that he or she is experiencing fear". Except in conditions that can be considered abnormal, there is a general consensus or opinion about how we express ourselves and, therefore, "in the scientific study of emotions it is not unreasonable to appeal to our intuitions about what emotional states are typically produced by situations of certain kinds". (ibid : 9) James may be revolving in his grave but, as they point out, there is nothing he can do to them now!

The theory they propose is based on two central questions: "What is the cognitive structure of the emotional system as a whole?" and "What is the cognitive structure of individual emotions?" and their "working characterization views emotions as valenced reactions to events, agents, or objects, with their particular nature being determined by the way in which the eliciting situation is construed" (ibid :12 - 13. - their italics). The diagram of their emotional types can be seen here

A further dimension they give to their analysis is that of the variables affecting 'intensity' of emotion. These variables may be few, Joy and Distress being only affected by "the degree to which the event is (un)desirable", or several, as in the case of Happy-for[4], for example.

This theory is decidedly more complex than the others hitherto considered. Averill(1989), reviewing the book, describes it as "a tightly reasoned, succinctly written, and intellectually demanding work" as well as being ambitious. He also says that a "good book always points beyond itself" and takes the opportunity to open a discussion as to whether the categories of Ortony et al. can be regarded as 'natural kinds', which, like mathematical entities, have some sort of 'essence', and are traditionally conceived as in the class which remains invariable over place and time, or 'logical individuals', which have no 'essence' , need not be all-or-none, and are localised in place and time. If one talks of 'basic emotions', he suggests, one is referring to some sort of 'natural kinds', whereas if one believes that emotions can evolve over place and time, one is referring to 'logical individuals'. Ortony et al.'s categories, he claims, are somewhat ambiguous on this count, although if the ambiguities were removed, it is probable that the taxonomy presented would be of natural kinds. However, he feels they "leave open the possibility that the features that define a category may change across cultures and over time, which is more a characteristic of logical individuals". Perhaps he has a point, but although I agree that the taxonomy is probably one of natural kinds, I suspect, as a multi-lingual, - and I am sure Wierzbicka would agree with me - that he may be showing a monolingual bias against the possibility that natural kinds may just be sufficiently basic to allow for subsequent varied interpretations across cultures and time. After all, Ortony et al. go out of their way in the first pages to stress this point.

These attempts at classifying the lexicon show how elusive the subject can be. However, several psychologists, including Fridja, who have doubted the advisability of even trying to worry about the lexicon, seem to be more open to the subject now than they were a decade or two back. As we have seen, Davitz is appreciated better now than he was then. The more linguistically orientated authors referred to, like Fillenbaum and Rapoport (1971) and Wierzbicka (1972) are perhaps the most discouraged, and discouraging, although Wierzbicka (1992) is more optimistic about the power of the Natural Semantic Metalanguage she discusses. Those analysing metaphors rather than traditional language concepts may find that this approach is suitable for some languages, but may encounter cultural difficulties in applying it to others. However, the two groups engaged in the battle of introducing their robots to the problem of emotion have, perhaps, given hope to those who believe an analysis of the emotions not only can, but should, be attempted.

Oatley and Johnson-Laird (1990) are primarily interested in emotion words in English, and have gone to a good deal of trouble to draw up an impressive lexicon. Their analysis is attractive, and rather more accessible than Ortony et al.'s, but it has its weaknesses. They claim that "the folk theory of emotions embodied in English converges with scientific theory in identifying categories of analysis". Well, which came first, the chicken or the egg? Does language, or the English language in particular, reflect scientific theory, or is 'scientific theory' constructed from the 'intuition' built out of language? After all, one of the weaknesses they recognise in their own theory and that of their rivals', is that what any of them find intuitively acceptable, may not be so acceptable either to their rivals or to the reader. Their emphasis is on the words used to describe emotion situations, although the tests they use with feel + emotion word in a sentence to justify the inclusion of particular lexemes in their lexicon is a rudimentary attempt to project the word into an Emotion context.

Ortony et al.'s view, which makes a distinction between the emotion situation and the emotion word, and sees the emotions as valenced reactions that can be defined according to 22 functional types, has a lot to offer a bilingual analysis as it is not limited by the English words assigned to these types. This proposal to analyse situations rather than words would seem to owe a certain amount to the influence of Wierzbicka whom they appear to have consulted. Conversely, although her approach owes much more to linguistics, she seems well aware of the attitudes of psychologists to emotion. Her aim - to find a way of analysing semantic universals which, although phrased in one language, probably English, is sufficiently rigorous, yet flexible, to allow for the analysis of word/concepts in any language - runs parallel to their attempts to isolate situations which, whether or not they are lexically realized in any particular language, can be seen to represent all possible varieties of emotion.

It is significant, however, that, instead of suggesting that either the concepts or the situations can be reduced to simple components related to physically experienced and expressed factors, both Wierzbicka and Ortony et al. prefer to analyse emotion situations in terms of 'scenarios'. This is an approach which owes much more to the type of cognitive psychology which holds that emotions are the result of complex cognitive processes which take into account not just the immediate sensorial input from the outside world, but also the way in which this input is both analysed and modified by the information stored in the brain.

These theories have been put forward by people with a background of linguistics, psychology and AI and, in their different ways, their aims are to do with both providing cognitive psychology with food for thought, and contributing to the formalisation of language for AI. Linguistically, however, some of their hypotheses leave a lot to be desired. Although they have gone beyond previous attempts, like Davitz's, to classify everything based on the noun form, and admitted different types of verbs and adjectives, I would suggest that, unless they are prepared to consider the syntax, which, ignoring the semantic content of syntax, they have presumably left to their more syntactically orientated colleagues, as well as the context in which these lexico-semantic structures occur, their views will continue to be limited.

Whether one is talking about emotion or the lexemes used to describe or categorize it, it is easy to forget that by the time anyone actually describes an emotion in a normal everyday context, the emotion situation has gone well beyond any appraisal, arousal or behaviour stage. If the Speaker is the Emoter, the speech act will be a conscious evaluation and description of the event, and if the Speaker is referring to emotions in others, it will be an evaluation of the behaviour of the Emoter, or even a report of what the Emoter has actually evaluated in relation to his/herself. The Speaker, therefore, has opted for a description which social and cultural factors have taught him/her to associate with whatever it is that has been felt, or observed in others.

When trying to communicate feelings, people often struggle with words, perhaps conveying more of what they feel through body language than articulate speech. In certain cultural situations, there is an effort to control rather than show emotion. The English have always been considered cold by Mediterranean people because of their restraint in emotional circumstances, although, as an expatriate, I have noticed that the English have become more physically demonstrative since the 60s - perhaps as part of the Love you hippy phenomenon. However, there is still a certain amount of truth in the observation that the Englishman's idea of showing sympathy at a funeral is a touch on the elbow, and I'm awfully sorry, old chap. Living in a bilingual situation, I have also found that Portuguese has a wider variety of suitable things to say, or formal phrases, in situations like funerals than English. A colleague, who has spent some years in Turkey, tells me that Turkish is even richer in such formulae than Portuguese.

The description of an emotion may not actually use an emotion word. As we have seen, Wierzbicka's definitions and the examples she chooses from Tolstoy avoid the use of the emotion word itself. Davitz's descriptions of the physiological and psychological feelings associated with emotions also deliberately avoid the use of emotion words, as does the approach which tries to describe these abstract words in terms of the physical metaphors that we use to explain them. As Davitz's study showed, however, these descriptions were highly subjective and there was often considerable lack of agreement on how best to phrase them. Although it is certainly of interest, and probably quite amusing, to collect the metaphors with which people describe their experience, I would suggest that this method might produce a highly subjective set of examples which reflected not just the imaginative capacities, or otherwise, of the subjects who produced them, but also the intervention of the researcher's notions of what was, and what was not, acceptable within his/her corpus of examples.

Some may see the use of Emotion words as restrictive, in that they impose the socially structured interpretation of each word on what is seen as an essentially fluid and confused area of individual human experience. However, here one could answer this with Lutz's (1987 : 307) claim that "the role of the emotion word is central to the storage and structuring of ethnotheoretical knowledge in this domain" and that "Emotion concepts have embedded in themselves crucial cultural propositions and in turn are nested in larger networks of knowledge about persons, roles and goals". The first approach may be all very well for those looking for the Holy Grail of semantic universals applicable to all languages, but the latter may be of more relevance to the understanding of the relationship between real languages and the cultures which use them.

Emotion words in context may well reflect an honest evaluation of the emotion situation, but they will also reflect what the Speaker wants the Hearer to know about it. A parent will say to a small child Mummy's not angry, darling, but why did you paint pictures all over her new wallpaper? Besides, there are quite a few phrases using emotion lexemes which we use, regardless of our real attitudes to the situation, for example when one politely says I'd love to to come your party next week, when one is secretly thinking What a bore! I'm afraid often has little to do with fear, expressing polite regret as in I'm afraid the train is five minutes late. Although it is hardly in our interests to teach our robots insincerity, many computers are already programmed to answer one with user-friendly social formulae. My computer greets me regularly with Bemvindo a Macintosh (Portuguese for Welcome to Macintosh ), although I do not need Searle to tell me that it is not aware of what it is saying!

One argument in favour of using literary texts to find examples of emotion lexemes in context is because it is in the interests of the author to provide descriptions of emotions and emotional behaviour that are convincing. The examples in these texts often take the form of direct conversation, and they contain many of the social phrases as well as insincere uses of the emotional lexicon, but that in itself is a point in favour of this argument. At least in these situations language is being used with the primary objective of communicating some sort of reality to the reader, and not obeying the intuitions of academics trying to make their examples fit their theory.

When an emotional situation is being described seriously, there appear to be three main ways in which emotion words, at least in English and Portuguese, function in these descriptions:

1 a) to describe the Subject, or Senser's, emotional state or processes, as in I love you, He is angry with you, or I feel depressed;

1 b) to describe the behaviour associated with the emotional processes or state, as in He looked terrified, She sounded upset or They waited anxiously;

2. to describe the qualities or behaviour of the Object of the emotion, or Phenomenon, as in He annoys me, She is irritating or the kitten is adorable.

The difference between 1 a) and b) is that the former assumes the existence of the emotion, whereas the latter is more tentative and merely describes the behaviour that might indicate it. Therefore, although emotion words do not describe the physiological sensations or the physically observable behaviour directly, they may be added to identify the emotion if it is not identifiable from the context, as in she trembled with fear or he screamed in terror. [5]

Wierzbicka, in both (1972) and (1992), does not confine herself to analysing one part of speech, choosing, a noun, adjective or verb form, according to which seems most appropriate. Johnson-Laird and Oatley have also chosen different forms of the lexemes, apparently on the same basis, but often with little regard to actual usage. For example, aggrieve and daunt are cited as causatives, when it is highly unusual to find anything but the aggrieved or daunted forms. This sometime upsets the focus of their examples.

My proposal to limit myself to an already extensive list of emotion words - 401 for English and 323 for Portuguese - is made with a view to attempting to understand how folk wisdom, or the "ethnotheoretical knowledge" described by Lutz is stored not only in the lexemes but in the type of syntax and pragmatic context associated with them. For this I believe it is more economical in the long run to use the 'scenario' approach of Wierzbicka and Ortony et al.

In order to do this I have tried to collect a large enough quantity of examples, using the techniques described in the Introduction - over 25, 000 - to allow some sort of idea of the value of the different lexical, syntactic and semantic patterns that emerge, and when a more pragmatic interpretation is required, I have always tried to bear in mind the particular psychological scenario behind each example. Thus, my analysis of the examples not only recorded the salient syntactic features found to be immediately associated with them. It also included a description of the nature of the perceived cause of the emotion, or PHENOMENON, which might be implicit in the lexeme, but was often only retrievable from the context.

Since every individual emotion situation can be conceived of as unique, any attempt at analysing such a field as the lexicon of Emotion must suffer from some form of reductionism. Although I am prepared to agree that words do have a certain culturally accepted meaning, I prefer to arrive at suggestions about that meaning by means of seeing how it is activated when it actually appears in a sentence in context, than by examining it in isolation. Also, after all the space I have dedicated to the Mind / Brain problem, and to the idea of cognitive processes as being not just those based on immediate physical perception but also on all the complex social, cultural and, arguably, genetic, information stored in the brain at a conscious, sub-conscious or unconscious level, it is unlikely that I would now try to explain the language of emotion in terms of componential lexicology.

Rather than adopt any one theory described above in its entirety, I propose to adapt Ortony et al.'s (1988) classification of emotion groups, and add to it when necessary. It is true that their Emotion types could be reduced to fit into one or another set of basic emotions, but I feel the distinctions they make to establish emotion situations rather than just words, or metaphors, are interesting, whatever the reservations I share with Wierzbicka about the actual phrasing of their definitions.

It is not my intention to define the various lexemes to be analysed - other people have already spent a lot of time doing that, and, for the time being, I shall accept their conclusions. In later chapters, each group will be examined in detail, both from a lexical point of view and in relation to the syntax and type of context in which they appear. Any remarks on the differences and similarities between English and Portuguese lexemes are based on observation of the corpora, and on unmarked usage, rather than examples which, for some reason or another, appear unusual.

Ortony et al.´s proposal of a somewhat complex set of Emotion types on the basis of the possibility of their existence in any language, rather than their actual existence in English, the language of the researchers, means that some involve very few lexemes, and some none at all. It also means that certain lexemes, in different syntactic structures or different contexts, may belong to different types. One should also remember that although the language may not lexicalise these types, they can usually be expressed at the level of the text.

The definitions of the scenarios Ortony et al. use are couched in language which is intended to be as neutral as possible and which, as they themselves admit, can sound a little incongruous, as when terror comes under the description of [BEING DISPLEASED ABOUT AN UNDESIRABLE EVENT]. However, they have not faced the problem of language sufficiently, for they seem to fail to realise how difficult it is to disconnect one's ideas from the language in which they are presented. I should like to draw attention to the fact that, as a linguist, I have to examine emotion words in context, rather than emotional situations. I would also suggest that if these emotion words have a particular meaning for their speakers, the psychologists trying to ignore part of this meaning do so at their peril. Also, if there is little or no lexical proof of the existence of groups like Happy for, one should not forget that if a language does not lexicalise something, it is because the culture which uses that language does not feel the need to do so.

Ortony et al assume that an emotion is some sort of relationship between the SENSER and some sort of reality, or PHENOMENON, which is cognitively assessed, although the consciousness of this appraisal varies in degree. They distinguish between these PHENOMENA as EVENTS, which are "simply people's construals about things that happen, considered independently of any beliefs they may have about actual or possible causes", OBJECTS which "are objects viewed qua objects", and AGENTS which can be people, but also "non-human animate beings, inanimate objects or abstractions, such as institutions, and even situations, provided they are construed as causally efficacious in the particular context" (ibid : 18). Their term EVENT is a little confusing at first sight in a linguistic analysis. It must be taken as elastic enough to cover not just 'happenings', but also behaviour, more stative facts about the Other's world and concrete objects.

For example, they classify Fear as an EVENT-based emotion. The psychologist can both refer back to consciously perceived events that led to this fear, for example the sight of the snakes in Mary was afraid of the snakes, or, with a bit of help from the psychoanalysts, to events which affected the subconscious, and forward to the hypothetical event (of the snakes hurting Mary), all of which will cause the SENSER to feel Fear. The linguist cannot get round the problem that easily, and has to explain She was afraid of the snakes in which snakes would be classified as OBJECTS. The most important difference between the linguistic analysis I have made and their more psychological one, arises over their distinction between EVENTS, OBJECTS and AGENTS, and my attribution of these causes of the emotion to the 11 types of PHENOMENON which will be discussed in Chapter 5.

Since Ortony et al. are cognitive scientists combining an interest in psychology with one in AI, it is natural that their analysis should show the influence of logic on the way they create their framework. This means that they end up with the tree-type diagram, based on the notion of emotions as 'valenced reactions' rather than a less-structured list of lexical fields. Like most attempts by logic to explain natural languages, theirs also is likely to run into difficulties, but the mental discipline such an analysis imposes can be beneficial, and I frequently found it useful as I battled with the somewhat less than logical behaviour of the traditionally anti-rational field of Emotion.

One of the principal problems that emerged by trying to impose this cognitive psychological framework on two different languages was that actual lexical usage does not always fit easily into the type provided. Pity and pena, for example, had to be broken down into those examples which favoured a Sorry For, or compassionate, interpretation, and those which were describing some form of Remorse. The examples for hope overflowed completely from its Emotion type into a more strictly intellectual area, and the word usually used in Portuguese to translate it, esperar, got tangled up with the notion of waiting for.

Apart from the problem of fitting lexical usage into one type or another, there was the added problems that some, although largely inside one group, had connotations which involved them with another. Joy and Distress, on the other hand, produced such a variety of lexemes that one is tempted to make subdivisions of the group. Some lexemes were only on the margin of Emotion, like peace, but were included for their association, in some lexical uses, with a particular group, in this case Joy. However, I do not see these as reasons for abandoning the attempt to use Ortony et al.'s framework, because in the final analysis, it did help a lot to distinguish between different lexical uses, and the system of variables they propose also helped to clarify certain points.

Ortony et al. begin by discussing the Fortunes of Self, or Well-being emotions, Joy and Distress. Next they postulate 4 groups of Emotion related to the Fortunes of Others, Happy for, Pity, Gloating and Resentment. Judging from the corpora, however, neither English nor Portuguese speakers frequently feel emotions for others, and at the level of the lexicon these types are small or non-existent. Happy For is not specifically lexicalised in English or Portuguese, and the English gloat has no corresponding lexeme in Portuguese. For these reasons the data on these areas is extremely limited. The Pity and Resentment types, however, contain a few lexemes which fit quite neatly into the classification suggested.

The next set of groups are described as the Fortunes of Self emotions and are considered to be "prospect-based". Hope and Fear which, according to Ortony et al., relate to the prospects of desirable and undesirable events, are the two generally recognized emotions which appear in this group, and they are seen as occurring prior to the other four. The other emotions result from the confirmation or disconfirmation of the circumstances surrounding the Hope or Fear emotions, and are classified under Satisfaction and Fears-confirmed, and Relief and Disappointment. Of these 4 areas only Fears-confirmed provided no lexical item on which to collect data, although no doubt non-lexically marked examples could have been found. However, given the methodology used, it was impossible to collect such contextual examples satisfactorily, and so this section is not given any specific attention.

The next 8 emotion types are partly related to the Event-based Well-being emotions, but are also influenced by the Attribution Emotions, and by the "degree in which the actual agent is in a cognitive unit with the self". I departed slightly from Ortony et al.'s listing here and combined Pride with Gratification, and Self-reproach with Remorse, largely because it was difficult to separate a rather sparsely represented area into separate groups when these groups shared most of the lexemes. The analysis of the resulting groups will, however, take the distinctions made by Ortony et al. into consideration. Appreciation is a large group and it is balanced by a group Ortony et al. labelled Reproach. The group for Gratitude was small but distinct enough to warrant individual attention, and Anger was one of the bigger groups overall.

The valenced reaction to aspects of OBJECTS naturally divides into the Liking and Disliking groups. Love and hate describe what are many lay people's notion of true emotions, but not all psychologists have felt the same. The way in which they are reactions to OBJECTS brings them closer to a consciously cognitive explanation, and there would seem to be a certain gradient of meaning linking Liking to Appreciation in one direction and Desire in the other. This point becomes clearer when the syntax is considered.

However, other lexical groups were formed that are not acceptable to Ortony et al., but are considered as emotions, or as very close to the field, by other theorists. One of these groups would seem to refer to more general aspects, and can be classified as the Generic lexemes, such as feeling, emotion, sentiment and passion. This group can be considered as a sort of metalanguage of emotion and as such should be considered separately from the directly referring lexemes. Each language groups is heavily dominated by one lexeme - feel for English and sentir-se for Portuguese - and the various problems this causes make it necessary to use different criteria for analysis than those used for the other groups.

Ortony et al.'s Emotion types will not suit everyone and two more categories will be examined here for this reason - Surprise and Desire. They prefer to discount the surprise / shock type of lexeme as not an emotion per se, but rather as a variable affecting the intensity of the emotion which follows it, something that is essentially cognitive in origin. Whereas Shock usually accompanies Fear or Grief, Surprise might be equally followed by Joy or Sadness.

Although the argument is sound within their theory, it could also be pointed out that, whatever the subsequent emotion, Surprise and Shock can be observed in terms of arousal and behavioural factors, and also analysed as being the result of rapid cognitive appraisal. Surprise is a favourite emotion of those who categorise Emotion according to its physical expression, particularly those, like Ekman, who emphasise the importance of facial expression. Besides, one cannot ignore the physical reaction associated with Surprise and its synonyms. Shock is even a medically accepted term for a physical state caused by emotional disturbance. Therefore, this lexical area cannot be ignored here. The syntactic data was not dissimilar to that found for the other Emotion groups selected by Ortony et al.

Ortony et al. (1988) talk about 'desirability' as a variable affecting events to which other emotions are related, but ignore Desire itself as an emotion. Within their theory of appraisal, which is fundamental to any cognitive theory of Emotion, they look upon the process of wanting or not wanting something to happen as some kind of evaluative or cognitive process which precedes emotion. If several types of information are being processed simultaneously in the brain, as many believe, one could, perhaps, argue that desirability is only the type of variable they describe. Oatley and Johnson-Laird, however, who excluded Desire in their first analysis, were considering including it to balance Disgust in their second.

It is true that Desire is not found in many more recent lists of emotion although, as Fridja (1986) points out, Spinoza considered it one of the three primary emotions. Fridja himself, however, classifies it as an emotion, albeit a somewhat cognitive one, and sees it as a "tonic reaction" with "action tendencies [that] aim at achieving satisfaction states". He further believes that "desire itself is an emotion proper", and presumably he is referring here to the lexeme desire, as in his desire for peace, rather than the lexical group of words classified under Desire. This is an example of the many occasions on which psychologists and other non-linguists produce confusion by being less than careful about their language. However, if one requires some kind of physiological or psychological tension as a criteria for defining emotion, Desire must be at least a candidate for consideration. It need not be sexual desire. Whatever the arguments in favour of the cognitiveness of the child's I want computer games for Christmas, or many people's I wish I were a rich man, one can hardly ignore the gut reaction that is often associated with them. However, since the number of examples collected for this area was considerable - about 3,700 - and the syntactic data would have distorted results on the other groups unacceptably, this group has been considered separately from the others.

Another area which I considered at the beginning was Courage, which is included in lists such as Arnold's, but its status as an emotion is very doubtful. It is related to Fear in that it describes an absence or control of Fear. However, the SENSER is not so much experiencing an emotion here, as cognitively summing up all the social and cultural traditions that he/she has internalised in order to deal with Fear. Whether what the SENSER then does can be called Courage will be decided not by the SENSER but by an external observer. For these reasons this group of lexemes was not included in this study.

The lexemes selected for each group were similar to those which appeared in the appropriate semantic grouping in the thesauri or dictionaries of synonyms consulted. The preliminary study in the Birmingham Corpus included an even wider variety than appears here but not all of them appeared in the EC. The selection of words for Portuguese were fewer in number. I recognize that this may be partly due to my lack of knowledge of the lesser known Portuguese words in this area, but the individual dictionaries formed from the different texts[6] were examined with care for new words, and few were forthcoming. However, since this is a subjective area of language and one which is used frequently in everyday life, it is only natural that English, with its tendency to absorb words from both the Romance and Germanic languages in these circumstances, should offer a wider variety of words[7]. It is therefore possible that Portuguese has a narrower range of words from which to choose, although, as we shall see, Portuguese shows a wider use of its more restricted choice than English, which tends to favour a small number of lexemes in practice.

Another difference between the languages is the tendency of Portuguese to favour stronger or more expressive lexemes, both qualitatively, i.e. offering a wider choice of lexemes in the more expressive range of the lexical group, and quantitatively, in the number of examples found. This is particularly true of groups like Joy and Distress, and possibly demonstrates a tendency of the Latin temperament to prefer stronger emotional language, similar to the tendency to favour stronger lexemes in relation to effects of Light, as shown in Maia (1988). The choice of these stronger words, both in an original text and in a translation, will vary considerably according to the situation and the taste of the writer, and this is one of the areas where the choice of word in translation is very delicate if one is to avoid over-statement, or a pompous or comic effect through inappropriateness.

The lexical nature of the groups under consideration will be examined in later chapters with a view to:

c) establishing the similarities and differences between the synonyms in the same language.

This analysis will refer generically to the type of syntax and context which helps to differentiate shades of meaning between lexemes, and even between uses of the same lexeme.

Since no lexical area is hermetically sealed off from from any other, decisions as to which lexemes were analysed in which group were based not so much on lexical considerations as on the nature of the PHENOMENON and the general situation which occurred most frequently in the examples. The divisions are by no means watertight and the inclusion or exclusion of certain minor items is sometimes, I realize, debatable.

Each lexical group will also be described in relation to the other groups considered and to other semantic fields when necessary, as the semantics of the different lexemes usually allows for elements from other emotions to be implied by them. My study of these areas in the much larger Birmingham Corpus (BC), it was possible to observe the 'company kept' by the different lexemes[8]. There were four main ways in which emotion lexemes seemed to appear with others, usually linked by and:

b) with antonyms, either other emotions, for example love and hate, or other notions, for example business and pleasure;

c) with words indicating emotions which are often associated together, for example love and happiness or fear and rage;

d) with words indicating factors that might contribute to the emotion, for example, health and happiness and pleasure and profit..

The data I managed to collect on this type of material was only really relevant with the more well-represented lexemes, but it frequently helped in their allocation to certain groups and in distinguishing between ones in the same group.

I hope these considerations will help explain my choices at a lexical level and serve as an introduction to the more syntactic and semantic analysis which is the subject of the following chapters.

[1] The percentages quoted after each cluster refer to the % of those questioned who associated the emotion with the cluster. For example, in describing anger, 53% of those questioned associated it with hyperactivation, but only 16,8% with inadequacy.

[2] In (1972), an example of the type of definition she suggests is:

X feels sad = X feels as one does when one thinks that what one has desired to happen has not happened and will not happen.

[3] Her definition of sad, for example shows more careful preparation than the one given in 1972:

Sad (e.g. X feels sad)

X feels something

sometimes a person thinks something like this:

something bad happened

I would want: this didn't happen [i.e. I wish it hadn't happened]

because of this, I would want to do something

I can't do anything

because of this, this person feels something bad

X feels like this

[4] Happy-for is governed by the variables:

1) the degree to which the desirable event for the other is desirable for oneself;

2) the degree to which the event is presumed to be desirable for the other person;

3) the degree to which the other person deserved the event; and

4) the degree to which the other person is liked.

[5] Examples of this kind were tagged in the corpora for BEHAVIOUR. These examples are always SFoc, as the PFoc lexemes, by definition, will describe the emotion-provoking behaviour or qualities of the Phenomenon, in any case.

[6] The software at my disposal, STABLEX by Professor André Camlong of the University of Toulouse, allowed me to list every word used in a text and the number of times it appeared.

[7] For example, felicity, although rarely used, can be found as a cognate for the Portuguese felicidade, but Portuguese has to use this word and others like alegre to translate the English glad, glee, happy and merry, all of Old or Middle English origin, according to the S.O.E.D.

[8] I am grateful to M.A.K. Halliday for the personal suggestion that I should study this aspect of the various lexemes when consulting the BC. The idea was no doubt related to his theory of 'lexical cohesion', and, although I only had time to observe those words which occurred in the concordance immediately before and after the lexeme being studied, the results were often illuminating.