CHAPTER 5

The aim of this chapter will be to discuss the semantic aspects of the Emotion lexemes. It is difficult to find one school of linguistics which deals satisfactorily with all the relevant aspects, so I shall be somewhat eclectic in my choice of theorists. I feel that much of the problem resides in the fact that the linguists involved are often conditioned by the tenets of Behaviourist psychology, and I hope to show the change in emphasis that occurs when one approaches the subject from the viewpoint of the type of Cognitivist who sees the mental processes as including both physical perception and complex brain / mental processes. This chapter will analyse the idea of Emotion as the relation existing between the SENSER, or individual who feels emotion, and the PHENOMENON, or whatever produces emotion in the SENSER, and the way linguists have tried to account for this using deep case theory. It will also look at theories that have been advanced for the semantic classification of verbs, and to account for the syntactic behaviour of the verbs of Emotion.

In the psychological theories of Emotion discussed up till now, the focus was on the SENSER and the emotion. The arguments centred on how the SENSER experienced the emotion - was it mental or physical? did mental recognition of the emotional state precede or follow the physical manifestations? were emotions innate or acquired? - were some of the questions asked. In fact, so much has been written about Emotion from the point of view of what happens to the SENSER, that the PHENOMENON tends to take a very secondary position of interest. The term PHENOMENON has been chose because to use any of the following terms - cause, reason, stimulus, target - amongst others, is to opt for one or another explanation of Emotion. Later, when examining the PHENOMENON in more detail, further classification of types of PHENOMENA will be necessary.

The SENSER of an emotion is necessarily a human individual, or group of individuals, although one may make such SENSERS out of animals, machines, or the characters of myths or science-fiction through a process of anthropomorphism. The PHENOMENON, on the other hand, can be virtually anything, and can be concrete or abstract, or even a figment of the SENSER's imagination. In fact, the PHENOMENON is essentially the SENSER's perception of some event, action, person, or thing. The same apparently objective reality may well produce violently opposing emotions in different people - politicians' actions produce delight in their supporters and dismay in their opponents.

Fridja (1986) probably speaks for most psychologists nowadays when he says that "Emotions are rarely, if ever, elicited by an isolated stimulus". Even then he claims to be using 'stimulus' as a convention to describe this phenomenon, and chooses the term 'antecedent' in more specific situations. He further defines between 'situational' and 'dispositional' antecedents, the former being immediately related to the specific emotion causing situation, and the latter being the various innate, cultural and other factors which dispose the person to 'emote'. Most emotional situations can, quite probably, be explained in terms of both the SENSER and this more multi-faceted kind of PHENOMENON.

A problem that has exercised people's imaginations is the so-called 'objectless emotion'. According to Kenny (1964), Descartes was worried about this, and Wittgenstein (1958 : 22), Freud (1953: 12) and others have discussed apparently non-directed emotions like angst. Although most people assume that an emotion relates the individual to something in the outside world, it seems to be possible to say things like I am happy / depressed / afraid but I don't know why. However, this does not mean, on further examination of the problem, that one cannot come up with a reason why. The very fact that the speaker says but I don't know why implies that he/she believes that there should be a reason, but that this reason is not immediately identifiable. With the help of someone like a psychoanalyst, the SENSER may be able to examine the emotion's antecedents for an explanation, and eventually point to a wide variety of insignificant and apparently unconnected stimuli which combined to produce the emotion.

Alternatively, there may be some biochemical disfunction or perhaps drug-taking which has induced the symptoms and feelings of euphoria or distress which the SENSER identifies as an emotion. Fridja also refers to the possibility that "emotions could occur for reasons of emotional metabolism or internal regulatory processes". Strong emotions rarely last long and often seem to trigger off the opposite frame of mind as a form of control. Most of us have experienced a feeling of deflation once the effects of euphoria have worn off, or a feeling of calm when our fury at something is spent.

Another explanation for the objectless emotion is to call it a mood. After all, the type of emotion being considered is usually one of the milder ones and more amenable to being analysed as the result of a series of diffuse stimuli over a longer period of time. However, it is difficult to imagine a context in which one could say I am angry / delighted / terrified but I don't know why, because these emotions would seem to require an immediate and identifiable object. In contexts where verbs like love and hate are used the linguistic structure requires a grammatical object in any case, and obliges identification of the psychological PHENOMENON. However, one might explain situations like I love her, but I don't know why using similar arguments to those put forward for the objectless emotion, because the syntactic verb+object structure can be seen as only specifying an independent sub-category of love at a deeper semantic level.

Emotion, therefore, essentially needs a SENSER and some kind of PHENOMENON. The relationship between the two is mediated by the emotion, but the emotion is interior to the SENSER. The nature of the PHENOMENON is also dependent on the SENSER. The Behaviourist view, that an emotion was the physical response to a particular stimulus, contributed to the interpretation of the PHENOMENON as the stimulus or cause of the emotion, and obscured the fact, defensible even within this way of thinking, that the PHENOMENON was not an independent reality, but was the SENSER's perception of some object or situation. This point becomes easier to grasp when one explores the role of appraisal in emotion put forward by the Cognitivists.

Some would say that the function of the PHENOMENON is to define the emotion. Thus, if John loves, that love will be understood only when we understand whether the PHENOMENON is his love for his mother, his girl-friend, his dog, or a cup of coffee. In many cases the PHENOMENON will need further definition. If one says John is angry with his wife, we may need the information that she has burnt a hole in his favourite shirt to better understand the emotion caused. Even then, the intensity of John's feeling can only be estimated if we know his normal attitude to his wife, as well as any attenuating circumstances, like his having had a bad day at work or his inability to afford a new one. Ortony et al. (1988: Chapters 3 and 4) have constructed a framework of variables with which to describe the various factors which may influence the intensity of emotions.

The question - Does one love because one is loving or because something is lovable? - was a problem Socrates and Euthypro discussed well over two millenia ago (Plato, 1954: 32), and it appears that nobody has really come up with a satisfactory answer since, for the debate continues. Most people would agree with Searle that emotion is internal to the SENSER and that one may love, hate, fear or be angry without anyone else necessarily being aware of the fact. However, one of the functions of emotion is to govern our relations with others, and for that reason it is more normal that the emotion should be expressed and observed. An individual's capacity to experience emotion is governed by innate, cultural and environmental factors, and some people are more predisposed to be loving for example, than others. For similar reasons, the individual also varies in his/her ability to appraise, control, show or dissimulate emotion. If one is loving, therefore, one may be temperamentally predisposed to love, or one may be actually feeling love for someone, or one may be demonstrating this love in one's behaviour, or even simulating love that does not in fact exist.

Whether an PHENOMENON is lovable or not is an appraisal made of the PHENOMENON by the SENSER, although it is arguable that one could objectively describe someone who one does not personally love as lovable. The PHENOMENON may or may not deliberately influence the SENSER's evaluation. John may be attracted to Mary without her even being aware of his existence - as, for example, when a cinema goer finds a film star attractive - but it is just possible that Mary, in order to make herself lovable to John, may follow all the traditional advice on how to achieve this effect. However, we are unlikely to describe this exercise in attracting John as being lovable, as the appraisal of whether Mary is lovable or not remains up to John[1].

Love, however, is one of the easier verbs to analyse, at least in English, because it is a verb in which the emoting SENSER is the Subject and the PHENOMENON is the Object, and what I shall refer to as a SENSER focusing, or SFoc, verb. Besides, the lexemes lovable and loving conveniently exist to allow analysis. One could analyse John's fear of spiders or war using similar criteria, and even work out examples to show the lesser or greater conscious appraisal involved, but fearing only exists as a present participle, not as an adjective, *fearable does not exist, and fearful is full of ambiguities as a central adjective related to fear, not to mention its frequent role as an intensifier, as described by Quirk at al. (1985: 7.87 - 90).

Many of the verbs with emotive connotations are PHENOMENON focusing, or PFoc, ones in which the SENSER is the grammatical Object and the PHENOMENON the grammatical Subject, as in the case of frighten or annoy, and this adds to the problems we are considering. The SENSER is now in an apparently Objective or passive relation to the verb's Subject. The examples John kicked Mary and John frightened Mary would suggest two similar events, and that in both cases John is the Agentive Subject and that he deliberately produces some sort of physiological effect on Mary so that, as the passive Object, she consequently suffers from either severe bruising and pain after the kick, or from the trembling, sweating and general gut reaction associated with being frightened. She might, poor thing, be subjected to both indignities and one could say, John frightened Mary by kicking her. However, one cannot say, *John kicked Mary by frightening her. This is because, whereas deliberately kicking someone is sufficient unto itself as an action, deliberately frightening someone requires performing some action which induces the frightened emotion in that person. Besides this, Mary's perception of the pain from the kick comes directly from the part of her anatomy affected. However, the physical symptoms associated with fear that she experiences can be explained either in James' terms, in which they constitute the emotion, or in Cognitive terms, according to which they are connected to her appraisal of the implications of John's kicking her - i.e. that he is aggressively inclined towards her and will probably proceed to hurt her further. Mary's being kicked depends on John's action, even if it is accidental, but her being frightened depends on her resulting arousal and appraisal of his action.

This becomes clearer if one considers that if John had intended to frighten her by kicking her, without realising she was a masochist, the fact that John thrilled Mary by kicking her is dependent on Mary's feeling this emotion despite John's intentions. It is also possible to say John frightened Mary without John's being minimally aware of the fact, in which case Mary's being frightened is entirely dependent on her. It is clear that kick and frighten cannot simply be classified together as, for example, agentive verbs. The examples in the corpora were tagged DELIBERATE when PFoc examples of any kind occurred explicitly in a situation when the PHENOMENON was deliberately trying to cause an emotion, but this only produced about 30 examples in both corpora - or about 0,12% of all the examples. Linguists trying to classify the type of verbs discussed above have found that the semantic analysis elaborated by deep case theory can help to explain the situation.

The notion of case is present at a surface level in many languages, and traditionally it was considered to be present in the form of suffixation or inflection of the noun form concerned. According to Crystal (1987: 92), Finnish forms fifteen cases in this way - although it therefore dispenses with prepositions - and Latin boasted six cases, but most European languages have dropped the suffix type case system in favour of prepositional phrases over the last few centuries. This is the situation for English, although pronouns still have accusative forms, and there is what is known as the ' 's genitive'. In Portuguese, pronouns have accusative and dative forms.

Case, with its connections with prescriptive linguistics and the tendency to try and force all languages to fit into categories originally devised for Greek and Latin, was an unpopular notion among linguists bent on describing languages such as English earlier this century. The deep case theory discussed nowadays does not often use the terms accusative, genitive and dative, perhaps due to a wish to avoid confusion between the surface structure grammatical category found in inflecting languages, and what is now considered a deeper level semantic role. It was as the notion of deep structure developed, that it soon became clear that "there was a case for case", as Fillmore (1968) explains. In this paper Fillmore brings together notions of case discussed by other linguists such as Tesnière (1959), Whorf (1965), Chomsky (1965), Benveniste (1966), Lakoff (1966), Halliday (1967), Bach (1968), and Lyons (1968) .

The deep case structure underlying verbs expressing Emotion is complex and controversial. This is not surprising if we think of the philosophical issues and psychological theory already discussed. It was soon recognised that certain verbs, those that Lakoff (1968) classified as 'stative', rejected agentive roles. Discussion centred round examples like I liked his attitude and His attitude pleased me. Fillmore (1968: 30) asserts that like and please are synonymous, with the frame feature +[___O + D], and differ only in their subject selection features. He also pointed out that like, "has in its history the subject selection possessed by please", a point that has been thought significant and has been discussed by others, including Allen (1986). It should not be forgotten, however, as Halliday (Kress (ed.) 1976: 164) pointed out, that both like and please have their own corresponding passive forms.

Modern case theory recognises that verbs can be divided into certain semantic categories, the most usual of which are active verbs as in John kicked Mary, in which John has an AGENTIVE (e.g. Quirk et al. 1984) or ACTOR (e.g. Halliday 1985) role and Mary is variously considered as having an OBJECTIVE (e.g. Fillmore 1968), AFFECTED (e.g. Quirk et al. 1984), PATIENT (e.g. Chafe 1970), or GOAL (Halliday 1985) role.

There seems to be fairly general agreement that one cannot talk of AGENTIVE or ACTOR roles in association with verbs like like, please, fear and frighten, but there seems to be more difference of opinion on how one does classify the semantic roles involved as can be seen from Figure 5.1

It is interesting how these terms reflect or reject current psychological theory on emotion. DATIVE, RECIPIENT and LOCATIVE, CARRIER (PORTADOR) or PATIENT emphasise the idea that emotions are something that happen to one, and reflect both the traditional attitude to emotion and Behaviourist theory.

Fillmore (1971) and Chafe (1971), with their proposal of EXPERIENCER, provide an interpretation which, although still recognising a somewhat passive role, suggests a measure of active recognition of whatever is felt. This term seems to have been quite generally accepted, perhaps because its slight ambiguity does not compromise the user to too strong a position on the problem.

Halliday, however, with his proposal of PROCESSOR in 1969, comes nearer the current psychological view of emotions. He refines this further in 1985 with his proposal of SENSER for the person who feels, or senses, and combines this with the notion of the verb concerned being a process verb, which is again nearer the Cognitive view of emotion. Sense can be used to imply both mental process and physical sensation.

Similarly OBJECT, PATIENT and AFFECTED roles are less satisfactory than PHENOMENON, CAUSE, or STIMULUS as explanations for the roles of S with please or O with like. It is all very well to argue a non-participatory role for picture in He liked the picture or the picture pleased him, but it is less satisfactory in cases like John frightened Mary, or threats in they feared his threats, they were frightened by his threats or His threats frightened them. If one insists on the OBJECT role for whatever causes the emotion, one must make the type of analysis made by Cook (1979: 141) in the following examples with their case frames:

and he explains that in the third example "the AGENT role is added to a psych movement verb, and is co-referential with the OBJECT stimulus," (ibid: 140).

Halliday (1969 and 1985) is very careful in his choice of PHENOMENON as it allows him to include things both animate and inanimate which may cause emotion either intentionally or accidentally. He probably avoided CAUSE as too vague a term, or because it could lead to confusion with Causatives, and STIMULUS may have smelt too strongly of Behaviourist laboratory experiments.

Helbig, with his elaborate set of deep cases chose the terms CARRIER and STIMULUS, but perhaps he was still influenced by a more Behaviourist view of the world. Quirk et al. (1984) have opted for the same RECIPIENT and AFFECTED roles that they put forward in 1972, and EXPERIENCER and OBJECT appear frequently in linguistic literature nowadays. However, I feel that Halliday's SENSER and PHENOMENON are perhaps the most suitable terms for a more cognitivist approach.

The data in the corpora confirm the notion that the SENSER must be either human or human-like. A distinction that could be made is that between the SENSER as identical to the Speaker, and the SENSER as observed or questioned by the Speaker. If one were to base one's data on a purely oral corpus, it would be interesting to see the relative quantity and quality of examples describing emotion which took this distinction into account. However, one would have to decide the ratio in which psychoanalysands' introspective monologues or the psychoanalysts' reports on the same monologues should be included in the corpus.

In a corpus taken from written texts, it would be even more difficult to make an analysis with more general implications properly. With novels written in the first person, as is the case with Great Expectations, one might attempt a more linguistic/stylistic examination of an individual writer's focus on the emotions, by assuming the Speaker to be identical with the character in the first person, and his/her views of others to reflect a reasonable interpretation of their emotions. Alternatively, one might try an analysis in which the main character's view, for example, that of Alice in Alice in Wonderland, and his/her observation of the emotions of others, are adopted as the Speaker's. These techniques, however, are not so easy to apply as might appear. They do not allow for material reported by others, for direct conversation in the text, or for various forms of author's licence[2].

Of more interest to my aims here is an analysis of the types of PHENOMENON that appear in context. As has already been pointed out, once anyone attempts to express anything about an emotion in language, a certain level of conscious appraisal is implicit in the situation. One factor which can indicate the degree of conscious appraisal involved, is the nature of the PHENOMENON to which the emotion is attributed. The way in which different lexemes, and groups of lexemes, require or reject different types of PHENOMENON can therefore be seen as an indicator of the consciously recognized nature of the emotion in question.

The PHENOMENON can be analysed by examining the syntax of the sentence. Some lexemes may imply certain types of PHENOMENON - for example, sympathy and compassion imply a feeling for another human being for some negative aspect of their lives -, and others may favour an idea of difficulty in attributing the emotion to a specific PHENOMENON - as with uneasiness which is usually the result of several ill-defined phenomena. Others require syntactic complementation which restricts the type of PHENOMENON available - for example, the verb pity does not allow (THAT) clause complementation, whereas fear does, and different forms of the same lexical item allow for more or less clarification of the situation.

If one takes the psychologist's viewpoint, a thorough analysis of the whole situation, or text, will eventually reveal the nature of the PHENOMENON. This will be obtained by an evaluation of the linguistically communicated information by the receiver within the limits of his/her knowledge of the world. However, the linguist must attempt to keep within the conventions of the language as much as possible, if he/she is to analyse anything in linguistic terms, and it is within these conventions that it is possible to classify the PHENOMENON as being:

There are also a few lexemes which have been classified as nominals, the nature of which will be discussed in Chapter 8.

Since the existence of a PHENOMENON is, however, presumed at some level, Type 1 will, in effect, absorb all those examples which are not explicable within the restraints of linguistic theory, and it accounts for 18%(E) and 21(P) of all the examples.

Type 2 is very limited, (0,9%(E) and 1%(P), but I felt it was necessary to account for those examples that refer directly to the Self, as in He was very satisfied with himself, or to qualities perceived explicitly as permanent, such as in He has a happy disposition or Ele é feliz por natureza. Although these examples represent different references to the Self, to further subdivide this already tiny group seemed unnecessary.

The third type, with 6,2%(E) and 5,1%(P), accounts for those examples in which the emotion can be attributed to the Senser's perception of his/her personal physical, psychological or situational state. More obvious examples of this type would be He is happy because he is rich, and He hates being in prison. However, it also includes examples which imply non-active, or stative, and RECIPIENT roles for the SENSER, as in He is happy lying in bed or he loves receiving presents.. Many of these examples include verbs described by Quirk et al. (1984 : 428 - 32) as verbs of stance as well as verbs of possessing and receiving.

Type 4 was chosen because I felt it was necessary to draw attention to the more perception based cognitive approach, and to distinguish between the more stative aspects of the situations accounted for by the previous group, and those introduced by verbs of perception or some kind of intellectual or mental process. This decision was encouraged by the fact that there was a fair number of examples such as I was sad to hear of his death or I love to see the sunrise , 4,9%(E) and 3,9%(P). This is a somewhat fuzzy area to work with as it calls into question the status of see as a perceptual verb or as a more active verb like watch. This is further complicated by the fact that Portuguese does not always make the almost obligatory distinction in English between see and watch, or hear and listen,[3] with the verbs ver and ouvir. The solution was to include all these examples in this group. It is a type that is significant because it would appear that, at a pragmatic level, there is a preference for examples like I was sad to hear of his death over I was sad about his death, and this naturally swings the focus away from the Other to the Senser's emotional or mental processes.

Type 5, with 10,3%(E) and 7,1%(P), includes all actions by the SENSER whether they refer to past behaviour - he was sorry for what he had done, habitual activities - he loves swimming, or future or hypothetical action - he is afraid of climbing the mountain tomorrow or he hopes to go to Italy next week. The timing of the action is often dictated by the emotion lexeme involved, as can be seen in the examples given, and usually requires expression in a clause.

The sixth group, the Other, turned out to be one of the largest numerically, with 19,2%(E) and 20,4%(P). In a way this is surprising because one might think that category 9, or actions by the Other, with 17,7%(E) and 15,1%P) would claim a lot of these examples. A broader psychological analysis would probably favour this balance, but the narrower linguistic criteria produced these results.

Types 7 and 8 are small groups, 2,6%(E) and 1,6(%P), 1,5%(E) and 0,6%(P), but were useful as a parallel to 3 and 4. The examples referring explicitly to the Other's physical condition or situation proved particularly useful for the rather frequent examples of the Other's death.

The tenth group, with 15,7%(E) and 21,9%(P), includes a wide variety of possibilities, as was to be expected, and at times it was difficult to decide between this and group 11, with only 2,6%(E) and 1,7%(P), the decision to attribute an example to type 11 being usually based on the fact that the situation was more completely expressed syntactically or pragmatically.

Whenever possible, the decision to attribute an example to a particular group was taken on linguistic grounds, and more often than not the PHENOMENON was either retrievable from the semantics of the lexeme, or from the semantics or syntax of the sentence and immediate co-text. Some were only retrievable once the conventions of reference, substitution and ellipsis described by Halliday and Hasan (1976) have been accounted for. A few were deduced through the kind of pragmatic convention known as Grice's conversational implicature, and some decisions were based on a wider knowledge of the text and situation.

One reason for keeping as close to linguistic evidence as possible is the danger involved in allowing one's own more psychological interpretation to interfere. It is, however, impossible to be completely objective. I should like to point out that my greater familiarity with English allows me a finer perception of the linguistic and pragmatic conventions implicit in the English texts than is the case for Portuguese, and I also admit that my greater knowledge of certain texts in the corpora may have influenced a few of my decisions.

This fact need not necessarily be considered negatively for it only goes to show the complexity of the relationship between language and life. It may, however, partly explain

why the PC has 3% more examples of type 1. The striking point, however, if one examines Table 5.1 closely, is how similar the statistics for both languages are for this area. Whether this is attributable to similar linguistic or cultural factors, or to some deeper more universal situation is very debatable. Only a thorough examination of several languages from different cultures could help to indicate a solution.

As Vilela (1992) demonstrates so succinctly, various attempts have been made by theorists of different schools to classify verbs, and each has been made from roughly one of four different points of view, which he classifies as basically syntactic, basically semantic, semantic + syntactic, and basically pragmatic.

The basically syntactic theories usually define between copular, intransitive, and various types of transitive verbs and accumulate information on individual verbs. The more sophisticated approaches specify the forms of complementation to be found with different verbs in considerable detail, as, for example, in the classifications by Gross (1975), which have influenced analyses made by Portuguese scholars like Malaca Casteleiro (1981) and Macedo de Oliveira (1979). Ranchod (1990) is also influenced by Gross, but her analysis of pronominal structures with the Portuguese verb estar is influenced by the theory which attempts to analyse verbs by comparing them to the paraphrases using 'dummy' verbs to which they correspond.

Valency grammar started from a sound syntactic base, but applied notions of semantics with a view to producing a theory which was both fruitful in syntactic rules, and yet economically organised according to semantic criteria. This type of classification of the verbs was applied in grammars of which Busse and Vilela (1986) is an example for Portuguese, and Allerton (1982) for English.

Vilela classifies as pragmatic the approach by those who, like Austin and Searle, are interested in the implications of the meaning of certain verbs for speech-act theory, are concerned with notions such as truth and the propositional value of language, and classify verbs according to certain pragmatic functions, such as asserting, evaluating, or reflecting the speaker's attitude. This is a development of the type of division of language into 'cognitive' and 'emotive' types proposed by the Vienna Circle.

Although I believe that it is impossible to separate syntax and semantics in any radical way, I shall obviously be using several notions of the more syntactically based type of classification in the following chapter. Here I shall be examining the way certain linguists have classified verbs from a more semantic point of view, often taking the syntactic behaviour of the verbs they classify as a basis on which to build their semantic classification. Halliday has always encouraged this process of analysis, and his contribution particularly significant.

One contemporary thinker who contributed to the semantic classifications of verbs proposed over the last twenty years is Vendler (1967: 110-112). He drew attention to the fact that the verbs like know or believe, which are usually classified as states, become processes when used with the progressive aspect. Vendler's point of view was philosophical rather than linguistic, but, as it became clear that semantic content depended on the coordination of lexeme meaning with syntax, linguists worked from this to produce categories for English. A distinction was made between 'stative' verbs, like those described by Vendler, and 'dynamic' verbs, which referred to most verbs which, by definition, implied some idea of action. The main arguments for defining stative verbs in this way were that they typically reject the progressive aspect and the imperative in modern English.

Comrie (1976: 48) describes the stative/dynamic division as "reasonably clear intuitively", although some may prefer to make a state/action or an event/process distinction. In relation to the stativeness of perception he also writes (ibid: 35) that "different psychological theories differ as to just how active a process perception is .......... [and] different languages are free to choose, essentially as an arbitrary choice, whether such verbs are classified as stative or not". The same comment, in a slightly more restricted form, could be extended to Emotion.

Dowty (1979: 66-7), within the field of generative semantics, classifies the stative verbs which are relevant to Emotion as:

These distinctions have affected analyses of the syntactic categories related to emotion. For example, the term psych-movement is used by Chomskyian theorists and fellow travellers, and much discussion has centred round the complementary relationship between the DI and D7 verbs and the D7 and E2 verbs. Dowty's aim was to classify verbs by examining their behaviour with aspect, and it is clear that the stative verbs do not like to co-occur with progressive or habitual action.

Other attempts at the semantic analysis of verbs have been made by linguists over the last 20 years, and Quirk et al.'s (1985: 201) categorisation is wel-known. They define between stative and dynamic types, the stative verbs being subdivided into quality, state and stance verbs, and the dynamic verbs being classified in terms of durative/ punctual, conclusive/non-conclusive and agentive/non-agentive. According to this theory, verbs are characteristically dynamic and adjectives usually stative. Thus, stative quality types are largely copula + adjective/noun structures, such as be irritating, and be a bore, or describe attributes as in have two legs. Stative State types are sub-divided into:

Verbs of Emotion typically belong to the stative state category b), although when John deliberately irritates Mary, irritate would be considered as an accomplishment, or dynamic verb which is conclusive and durative.

One point of interest which emerges from the categories described for emotion in all these theories is that they appear alongside those for intellectual states, perception and bodily sensation. It is significant that the syntax and semantics of these lexical areas seem to have a certain amount in common, and that they seem to be interconnected in a network of gradients of meaning and usage.

However, Halliday (1985: 101) does not draw the clear distinction between lexeme and syntax made so far, but prefers to discuss the problem we are analysing here as the "clause as representation" or in its "ideational function". He suggests that "our most powerful conception of reality is that it consists of 'goings-on': of doing, happening, feeling, being" and that "these goings-on are sorted out in the semantic system of the language and expressed through the grammar of the clause". He goes on to describe his idea of transitivity which "specifies the different types of process that are recognized in the language and the structures by which they are expressed".

Halliday, however, is very careful to avoid the stative/dynamic dichotomy and he prefers to describe the function of the verb phrase in the context of the clause as being related to different types of process. Although, in Halliday (1985), he does not develop the theme of "material process" to any great extent, his treatment of "mental processes : processes of sensing" (ibid: 106-112), which he sub-categorises as perception, affection and cognition, outlines most of the problems inherent to the lexical field of Emotion. He also draws attention to "behavioural processes" (ibid: 128-9) or "processes of physiological and psychological behaviour" which are grammatically "intermediate between material and mental processes". Verbs like look and think, as in Don't look! and I'm thinking, come towards the 'mental' end and are processes of consciousness that are being represented as forms of behaviour".

As we shall see in the next section, he associates these processes with specific participants, which others might refer to as deep cases. He also draws attention to the human nature of the SENSER with mental processes, and the wide variety of possible PHENOMENA. His attitude to the progressive aspect is that, although not impossible with these verbs, it is definitely a marked and unusual form. One further point that he makes is that the pro-verb do is not usually used with mental process verbs.

By avoiding the term of stative, with its static and passive connotations, and by categorising affection (or emotion) within a broad class of mental processes, Halliday's approach is more in line with current cognitive psychological theory than those of other theorists. Also, by avoiding strict definitions into fixed classes, and explicitly recognizing some form of continuous gradient between physiological and psychological processes, he allows for a more subtle semantic analysis of all these lexical areas than that afforded by the syntactically biased approaches offered by more didactically orientated theorists such as Quirk and his team. Without forgetting the wider semantic view, let us now turn to the information to be obtained from the syntax associated with the verbs and adjectives of Emotion.

5.3.2 The SENSER / PHENOMENON focusing distinction, and the classification of verbs of Emotion

The division of the lexicon of Emotion into SENSER and PHENOMENON focusing lexemes is clearly semantically based. Naturally, the theory of semantic roles of the participants can be coordinated with theories on the syntactic behaviour of the verbs. Others have described the two verb types, which appear to exist in most languages, although they vary as to their relative importance. As Talmy (in Shopen (1965: III: 99) points out, using the terms EXPERIENCER and STIMULUS:

He also points out that English used to favour STIMULUS type verbs even more and, as others have also demonstrated, like functioned earlier with a STIMULUS subject. The S.O.E.D. quotes an example as late as Wordsworth:

Both English and Portuguese have a limited number of SENSER focusing verbs, although there is wider variety of PHENOMENON focusing verbs. Despite the restricted variety of SFoc verbs, however, they are more important quantitatively, and, as Table 4.2 shows, only 5%(E) and 9,5%(P) of the examples were of PFoc verbs whereas SFoc verbs accounted for 19,2%(E) and 14,1%(P). This is particularly true for the EC where the difference is of 14,2%(E) as against one of 4,6%(P). It is also significant that different emotion groups vary considerably as to the percentage of SFoc or PFoc verbs they include. The Hope group in the corpora, for example, has only a few rather doubtful examples of excite to show as PFoc here, and the examples for the Desire group are also minimal. Other groups, like Anger, Gratitude and Joy have few or no examples of SFoc verbs.

However, if we add to these percentages those for the PFoc adjectives, the percentages change again. Many of the PFoc adjectives in English, for example, are formed by adding -ING to the infinitive form, as in alarming, depressing, irritating and disgusting. In Portuguese, one also finds adjectives derived from PFoc verbs ending in -ante/ -ente, as in alarmante, deprimente and irritante. These adjectives emphasis the SENSER's evaluation of the emotion-provoking quality of the PHENOMENON, rather than the latter's active involvement in the emotion. In the EC these adjectives outnumbered the PFoc verbs by 2 to 1, 10%(E) as against 5%(E), but the PC showed that the PFoc verbs and adjectives together carried almost equal importance, 9,5%(P) and 9,9%(P) respectively.

However, the combined percentages of PFoc verbs and adjectives still only account for a relatively small proportion of the overall lexical/semantic scheme of the different groups, 15%(E) and 19,4%(P) and a fair number of the adjectives included here are not even related to PFoc verbs. Why, then, has so much attention been given to the PFoc verbs? It is because some of lexical groups show high proportions of SFoc past participles that are derived from these verbs, as in alarmed, satisfied, annoyed and alarmado, satisfeito and zangado, 8,8%(E) and 11%(P), not to mention related nouns like alarm, satisfaction and annoyance, or alarme and satisfação.

Although the difference between SFoc and PFoc verbs and their related adjectives, participles and nouns is much the same in English and Portuguese, the two languages differ in the way they use certain structures within these patterns. The possibilities of phrase-making with two lexemes, with similar meanings in both languages, show that, although there is only one small difference in the first set of phrases, the second shows more:

Neither of these lexemes allows ter + noun, or estar com + noun constructions in Portuguese, but they should also be considered. Before moving on to an analysis of the problems of how these constructions influence the data on the corpora, it is necessary to consider the SFoc / PFoc distinction in relation to lexemes which do not have such a distinction built in to the morphology of the lexical form.

A fair number of these lexemes can be described as adjective/noun pairs with no related verbs. Perhaps one of the best examples of these are happy/happiness, and feliz/felicidade. Common sense analysis will suggest that happiness and felicidade. are emotions and that those people who experience them are happy or feliz. There is a tendency for these adjectives to appear frequently in sentences like he is happy or ele é/está feliz which allows for their classification as SFoc adjectives.

However, one sometimes finds that these adjectival forms appear qualifying something, usually an event or a mental process, which made him/her happy, or as qualifying the behaviour of someone presumed to be happy. We therefore find a happy day, um dia feliz, a happy memory, uma lembrança feliz, and a happy smile, um sorriso feliz. In the cases which describe something which made the SENSER happy, an PFoc classification has to be applied, although it is recognized that the adjective is not describing any essential quality of the PHENOMENON described, because the same event, objectively analysed, may have happy consequences only for the SENSER. What was a happy day for the winner of a race would be viewed quite differently by the losers. This demonstrates quite clearly how emotion-provoking qualities are projected by the SENSER on to the PHENOMENON.

In these days, when it is recognized that any analysis of an event or behaviour is dependent upon the observer's (Speaker's) subjective interpretation of the event or behaviour, I suppose one must accept that the Speaker's evaluation John gave me a happy smile does not necessarily imply that John was happy. However, common sense and good faith will normally allow one to interpret a happy smile as meaning that the person who smiles is happy. On this basis, these examples have been classified as SFoc, but have also been marked in the corpora for BEHAVIOUR. Other examples marked for BEHAVIOUR are the SFoc adverbs and examples in which SFoc nouns appear in adverbial phrases describing behaviour, as in an expression of happiness, and the average percentage of behaviour examples in the different groups is 15%(E) and 13,8%(P), as can be seen in Table 4.3. However, percentages vary considerably, from between 2,3 to 40,3% in the different groups, for reasons which will be analysed in due course.

On certain occasions it is difficult to classify nouns as SFoc or PFoc. In the example:

the position is not quite so easy. Although source takes the immediate PFoc analysis, irritation is still related to the neighbours, or SENSERS, by the preposition to, with its causative implication, and this prevents a clear SFoc analysis. On these occasions decisions were taken on an ad hoc basis with reference to the context. I cannot claim to have solved every problem satisfactorily, but since only a tiny minority of nouns are affected, any misjudgements will hardly upset the general balance of the statistics in the corpora.

[1] One cannot compare being lovable to its apparent Portuguese cognate ser amável because amável corresponds more to the idea of consciously being pleasant or agreeable, and is therefore used more in the context of behaviour.

[2] I carried out this type of analysis on texts early in my research, and concluded that one could use it to analyse individual tendencies to express or describe emotion. However, any results would be of more psychological, or literary, than linguistic interest.

[3] In English, see and hear , and other verbs of perception like taste, typically refer to what is seen as involuntary perceptual process. Watch and listen describe the voluntary use of the perceptual processes. This distinction is not always made with ver and ouvir, as can be seen in the examples Está a ver o jogo na televisão (He is watching the game on television) or Está a ouvir as notícias (He is listening to the news) . This type of example is common, although the verbs olhar and escutar exist to translate the deliberate use of perception expressed by watch and listen to.