Summary #e paper investigates the problem of semantic roles within the concept of verb-descriptivity (Snell-Hornby 1983). Descriptive verbs are semantically complex lexical items, where the modifying components are more focal than the verbal action itself (as in bustle or strut), and where the participants, the background situation and the attitude of the speaker emerge as distinctive elements (as in grovel and waft). As against orthodox views in early case grammar, a distinction is made here between the Patient as “sufferer” of the verbal action and the Vehicle as its “conveyer”. It is argued that this differentiation is essential for the understanding, the analysis and the translation of descriptive verbs, as the semantic roles are by no means identical when compared in various languages (here English and German). #is is illustrated by the comparison of lexemes in the semantic field of verbs expressing anger. #e aim of the paper is to sharpen awareness of such fine distinctions, particularly in their relevance for translation. Key-words: case grammar, semantic roles, verb-descriptivity, modificant, agent, patient, vehicle. Povzetek Članek obravnava problem udeleženskih vlog znotraj koncepta glagolskega opisa (Snell-Hornby 1983). Opisni glagoli so semantično kompleksne leksikalne enote, kjer so modifikacijski elementi bolj v žarišču kot samo glagolsko dejanje (npr. bustle ali strut) in kjer se udeleženci, ozadje situacije in odnos do govorcev kažejo kot razločevalni elementi (npr. grovel in waft). V nasprotju z ortodoksnim razumevanjem zgodnje sklonske slovnice, razlikujemo med vlogo Prizadeto, ki jo glagolsko dejanje spreminja, in udeležensko vlogo Posredovalec, ki je prenosnik glagolskega dejanja. T rdimo, da je to razlikovanje bistvenega pomena za razumevanje, analiuzo in prevajanje opisnih glagolov, saj udeleženske vloge nikakor niso enake, ko primerjamo različne jezike (tu angleščino in nemščino). T o trditev ponazarjamo s primerjavo leksemov iz semantičnega polja glagolov, ki izražajo jezo. Namen članke je izostriti zavedanje fines še posebej za potrebe prevajanja. Ključne beside: sklonska slovnica, udeleženske vloge, opisni glagoli, modifikant, vršilec, prizadeto, posredovalec. DOI: 10.4312/elope.3.1-2.127-139 #is contribution to the Festschrift for Meta Grosman goes back to the days when I was still working in lexical semantics, and for some reason, now forgotten, the essay never seems to have been published in this form. Many of the ideas and concepts so hotly debated at the time have meanwhile however proved invaluable for T ranslation Studies, and the contrastive analysis of German and English descriptive verbs should still be valid. #is contribution comes with the wish that someone might investigate the same phenomena for Slovene – or other Slavonic languages. Both in traditional grammar and in modern linguistics it is the accepted function of the lexical verb to denote a state, event or process (here subsumed in the phrase “verbal action”), while the function of modifying is conventionally ascribed to the adjective or adverb. #e subject of this essay is a type of verb which itself modifies the action it expresses: in other words, beyond its grammatical and syntactic functions, beyond the basic semantic situation created by the participants in the action and the circumstances accompanying the action, it contains a further focal semantic element that assumes the function of, and is expressed in the verb’s basic definition by one or more dynamic adjectives or manner adverbs. #e verbal action is in fact not merely stated, but – in the traditional sense of the world – described; and for this reason the verb concerned has been named the “descriptive verb”. #e linguistic phenomenon itself can then be termed “verb-descriptivity” (Snell-Hornby 1983). #e descriptive verb (DV) must be understood as a semantic complex basically divisible into a core of verbal action – or act-nucleus – and the descriptive, modifying element or complex of elements, which we may call the modificant 1 . An obvious example is the English verb strut, as defined in four standard English dictionaries: 1. walk in a stiff, self-satisfied way (ALD) 2. walk in a pompous manner, swagger (CED) 3. walk with pompous or affected stiff erect gait (COD) 4. to walk with a proud lofty gait and an erect head; esp. to walk with pomposity or affected dignity (Webster 3). 2 From these definitions the act-nucleus clearly emerges as walk, and the accompanying adverbial phrases express the modificant. When it is applied to isolated groups of lexemes, such a distinction is of course nothing new in semantic studies; in our approach however the DV is identified as a specific semantic structure independent of any onomasiological criteria, upon which most studies in semantic fields are based. Within this semantic structure, the verbal action itself is either relatively clear-cut, hence easily isolatable and itself already lexicalized (as in strut), or else it is only vaguely definable (as in English bustle or German schmieren) or even irrelevant (as in dawdle and trödeln) 3 , while the modificant is always focal, usually clearly definable and often highly complex. It is at the same time a part of the verb that only receives inadequate attention in most dictionary definitions. 4 #e modificant may express modality of action (loudness in shout and schreien, stiffness in strut and stolzieren), thus characterizing the verbal action directly; this semantic element, with which we are not concerned here, may be called direct descriptivity. More usually however the modificant characterizes and evaluates one or more of the participants (such as the agent, who in strut is self-satisfied and pompous), or it may specify certain background circumstances, such as the cause of the action (emotion or pain in scream, anger in toben) or the hierarchical status of the participants (as in grovel or kuschen); this semantic element, with which we are immediately concerned in this essay, may be called indirect descriptivity. Frequently the modificant combines both types of descriptivity: in strut, for example, the action is stiff and the agent is self-satisfied. As indicated above, the modificant of a descriptive verb consists largely of dynamic adjectives susceptible to subjective measurement, 5 thus expressing a distinct attitude of the speaker; this we may call speaker-evaluation. Most DVs, especially those denoting human behaviour, express negative speaker-evaluation, ranging from criticism of the action (as in meddle and jammern) to rejection of the agent (as in grovel and keifen). Furthermore the dynamic adjective, in being gradable, permits variations in degree and measurement on a scale; hence its value is relative and can be judged in terms of an implied and accepted norm. #is norm can be set personally by the speaker, as is his/her attitude towards speed (plod and bummeln refer to action felt to be extremely slow), or his/her perception as of movement or sound (whereby dash and sausen are experienced as being excessively fast, roar and krachen as excessively loud). Another type of norm is that set by the social environment and implicitly accepted by the speaker; usually this involves basic aspects of human behaviour such as speaking, walking, eating and drinking, whereby the concept of norm merges into that of speaker-evaluation, to indicate, for example, greed or “bad manners” on the part of the agent (gorge, schmatzen) or high consumption of alcohol, typically with the speaker’s professed disapproval (booze, saufen). 6 #e modificant consists further of merging, interdependent components of varying stress – as against the principle of clear-cut binary opposites in formalized componential analysis – whereby the focus can again be placed on modality of action or on one of the participants or attendant circumstances. #us focus is a further important concept in verb-descriptivity, and can well serve to differentiate lexemes of otherwise apparently identical semantic content: scream and yell, for example, differ in that yell focusses on sound volume and scream on the expression of emotion or pain. #e same principle can apply contrastively: while glow and glühen both include elements of light and heat, glühen focusses on that of heat and glow on that of light. Descriptive verbs can also focus on a participant, most typically the agent, or on one of the background circumstances; the near-synonyms blunder and bungle differ in that blunder focusses on modality of action, while bungle focusses on the result. #e difference between focal and peripheral components emerges for example from the grammatical forms most naturally taken by the verb concerned: present participle for focus on agent or action (“ein keifendes Weib”, “he went blundering on”) as against past participle in a passive construction for focus on patient or result (“der geplagte Bundesbürger”, “the affair was badly bungled”). Furthermore, while focal components are obligatory and cannot be contradicted (as seen in the unacceptable collocations *he dawdled quickly, *es glühte kalt), peripheral elements may be optional, and it is a literary device to use them deviating from their accepted usage. In postulating this semantic structure of descriptive verbs, we are of course indirectly concerned with various important issues in linguistics, among them componential analysis and case grammar. It must be stressed that the categories and distinctions outlined above emerged inductively from empirical methods and from direct observation of the usage of DVs, 7 and do not attempt to apply any specific linguistic theory. #ere is however a striking resemblance between these categories and those presented firstly by T esnière and secondly by Fillmore and other case theorists. Firstly, we have endorsed T esnière’s concept of the verb as the central element in the sentence, “un petit drame” (1969, 102) that presupposes an entire situation involving a set of participants (actants) and circumstances (circonstants). T esnière explicated his concept of the verb as follows: Le verbe exprime le procès. (…) Les actants sont les êtres ou les choses qui (…) participent au procès. (…) Les actants sont toujours des substantifs ou des équivalents de substantifs. (…) Les circonstances expriment les circonstances de temps, lieu, manière, etc. (…) dans lesquelles se déroule le procès. (…) Les circonstants sont toujours des adverbes (…) ou des équivalents d’adverbes. (1969, 102-3) A similar distinction is made by Halliday (1970, 149): In English, typically, processes are expressed by verbal groups, participants by nominal groups and circumstances by adverbial groups – the last often in the form of prepositional phrases. It is interesting that the syntactically realized categories described by Tesnière and Halliday are in part precisely reflected as built-in roles within the structure of descriptive verbs, whose participants are likewise expressed as modified and evaluated nominals.. While the circumstances, as explicated by Halliday and Tesnière, are only partially equivalent to those included in the basic definitions of DVs, there is a certain amount of convergence: in the statement “#e child screamed”, for example, the emotion inherent in scream (as “in terror” or “in pain”) would be realized in the basic definition of the verb by an adverbial phrase. In short, inherent in the semantic components of DVs are the same structures as those realized syntactically at sentence level: in its microstructure and specifications of semantic roles with their additional elements of evaluation and characterization, the descriptive verb has indeed the elements of a “miniature drama”. Secondly, in quoting the role of the agent in DVs, we have already implicitly entered the discussion, so intensely pursued by Fillmore (1968, 1971, 1977), Halliday (1967, 1968, 1970) and other theorists as to the type and definitions of the roles in case grammar. It is here that some basic discrepancies begin to emerge. a. Agentive and vehicular verbs #e broadest distinction made between the roles in DVs is that between the participants in the action and the circumstances accompanying the action, as delineated above. #e circumstances include the cause of the action (emotion or pain as in scream), the purpose of the action (as in haggle or drängen), the result (as in shatter or quetschen), prescribed surroundings (such as open country as in hike), and a possible hierarchy among the participants (inferior status of the agent as in grovel or flehen). #e participants on the other hand are involved in the action itself. #e central participant is the agent, the “one who acts”, the role which in our definition corresponds to Fillmore’s concept of the instigator of the action (1968, 24) and more precisely to Gruber’s definition of the agentive verb: An Agentive verb is one whose subject refers to an animate object which is thought of as the willful source or agent of the activity described in the sentence. (1967, 943, emphasis added.) Proceeding from this definition, one can classify DVs according to the role of their grammatical subject. Some DVs, typically those denoting human activities (eating, speaking, working, walking etc.), take an animate agent who instigates or performs the action as specified above. Typical examples of agentive DVs are kick, scamper, shout and scribble in English, and grinsen, plaudern, schleudern and schreiten in German. Other DVs, frequently in intransitive usage and with monovalent character, have as grammatical subject an often inanimate vehicle 8 , by means of which the action is manifested. Such vehicular verbs typically denote the perception of sound, light, and in certain cases movement. Examples of vehicular DVs are crackle, shimmer and waft in English and rattern, glühen and rutschen in German. It is with the role of vehicle that we depart radically from early studies in case grammar, which would include the vehicle under what is here called patient. #e empirical investigation of descriptive verbs has however shown the rigid distinction of agent and patient to be inadequate, as it forces the patient-role to assume what in verb-descriptivity emerges as two clearly differing functions. I would therefore replace the agent-patient dichotomy by a threefold spectrum or cline embracing the concepts of agent, vehicle and patient. #e agent is, as defined above, the animate, active instigator or performer of the action; the vehicle is the inactive and often inanimate “conveyer” through which the verbal action manifests itself; the patient is the passive undergoer or “sufferer” of the action, the being or object on whom or on which the action is performed from outside, and this normally presupposes the existence of an agent or instrument acting in complementary role opposition. #us in surface realization the patient normally appears as the grammatical direct object of an active (agentive) verb or as the grammatical object of a passive verb. #e conceptual distinctions between agent, vehicle and patient can be illustrated by the following examples: 1. $e soldier was pacing up and down. (Agent) 2. a) $e sun is shining. (Vehicle) b) $e river was flowing fast. (Vehicle) 3. a) #e sadist tormented the prisoner. (Patient) b) #e salesman swindled the old lady. (Patient) In the sentences under (2) it is clear that nothing is actually being performed on the respective grammatical subjects the sun and the river in the same way as with the prisoner (3a) and the old lady (3b); on the contrary, they are the natural phenomena essential for the manifestation of the verbal action. #e distinction emerges even more clearly if we convert the sentences into the passive: 4. a) *the sun is being shone. b)*the river was being flowed. 5. a) #e prisoner was tormented. b) #e old lady was swindled. 4 (a) and (b) are clearly anomalous; the verbs are in vehicular usage where the role of patient is not admitted. In 5 (a) and (b) on the other hand, the prisoner and the old lady assume the patient role of an agentive verb. #e conceptual core of the terms agent, vehicle and patient should now be clear; it should however be stressed that the borderlines are by no means clear-cut, and in practice overlappings are not uncommon. Much is obscured – or clarified – by the surface forms of the particular language under discussion: semantic roles, despite their postulated universality, 9 actually owe much to the fact that they were conceived on the basis of the English language and are invariably illustrated by isolated and sometimes idiosyncratic sentences devoid of any context. 10 Further confusion has been caused by the phenomenon of ergativity, which in English often remains undifferentiated in the surface realization of the sentence. #is is amply illustrated by some of the much-quoted sentences with which Fillmore opened the discussion of semantic roles (1968, 25 ff.): 6. a) John opened the door. b) #e door opened. c) #e key opened the door. d) #e wind opened the door. For further illustration let us also take the following sentences: 7. a) #e man is burning the wood. b) #e wood is burning. c) #e little girl dangled her legs from the chair. d) Her legs were dangling from the chair. John (6a), the man (7a) and the little girl (7c) all clearly emerge as the agent of the verb concerned. By common consensus of the early case theorists, the door (6b), the wood (7b), and her legs (7d) would normally be classified as the patient. In our definition of semantic roles, as the perspective of these sentences does not prescribe an agent or instrument acting in complementary role opposition, they are classified as the vehicle – although they do represent the point where the concepts of vehicle and patient begin to merge. Contrast however the following two sentences, where by virtue of the passive construction the perspective of the sentence implies the existence of an agent; 8. a) #e door was opened. b) #e wood is being burnt. In the examples quoted under (6) and (7) we can thus see that some verbs can be both agentive and vehicular, depending on usage and context. #is is however by no means always the case; as we saw in (4a) and (4b), vehicular and agentive usage is not freely interchangeable, and the construction used in (6b) and (7b) is by no means universally applicable. A similar grammatical transformation of (3a) and (3b) for example would produce the following anomalous constructions: 8. c) *the prisoner tormented d) *the old lady swindled. #e sentences (6c) and (6d) quoted above were in the 1970s the subject of much discussion, and it is not my purpose here to revive the debate. I would however like to point out that the surface similarity of these two sentences is in fact deceptive, and the problems they give rise to are again related to specific idiosyncrasies of English syntax. It is commonly agreed that the key in (6c) has the role of instrument, but it should be emphasized that here the usage of open is deviant, denoting unlock as against the action of causing movement involved in the sentences (6a), (6b) and (6d), and hence it should be discussed on a different semantic basis. Equally idiosyncratic is the usage in (6d), a sentence used by Huddleston (1970, 504) to illustrate the concept of an inanimate Force with the same function as an animate agent. Here again however open is not used in the same agentive sense as in (6a), whereby an act is consciously and wilfully carried out; the sentence rather means, in less artificial and more idiomatic expression, “#e wind blew the door open”, where, as in the two sentences (4a) and (4b), the wind emerges as vehicle of blew and the problem of an inanimate agentive Force does not arise. #e above arguments are strongly supported by the German translations of the sentences concerned, as we see from the following representative cross-section: 1. Der Soldat schritt auf und ab. (Agent) 2a. Die Sonne scheint. (Vehicle) 3a. Der Sadist quälte den Gefangenen. (Patient) 4a. *die Sonne wird geschienen 5a. Der Gefangene wurde gequält. 8c. *der Gefangene quälte. #ese sentences reflect precisely the same principles as we have seen in the English examples: the distinction between agent, vehicle and patient – as in (1), (2a) and (3a) – and the non- acceptability of both the vehicle of a monovalent verb in a passive construction (4a) and the patient of an agentive verb in an active construction (8c). Let us now compare German renderings of the remaining examples: 6a. Hans öffnete die Tür. 6b. Die Tür öffnete sich. 6c. ? Die Tür wurde mit dem Schlüssel geöffnet. ? Jemand schloss auf. ? Der Schlüssel passte zur Tür. 6d. ? Der Wind drückte die Tür auf. 7a. Der Mann verbrennt das Holz. 7b. Das Holz brennt. 7c. Das Kind ließ die Beine vom Stuhl baumeln. 7d. Die Beine baumelten vom Stuhl herunter. 8a. Die Tür wurde geöffnet. 8b. Das Holz wird verbrannt. Here three significant points emerge. Firstly, in contrast to the English sentences, ergativity is realized by distinct grammatical markers – in (7a) and (8b) by a prefix, and in (7c) by an auxiliary verb. 11 Secondly, the vehicular usage of an agentive verb is naturally realized in German by a reflexive construction (6b). #is illustrates a basic difference in the function of the reflexive in German and English; in English the verb is essentially agentive with a co-referential patient, viz. “He cut himself”, but not *the door opened itself. #irdly, the semantic haziness of the two problem sentences (6c) and (6d) is reflected in the confusion of the three German speakers asked to translate them, in the case of (6c) resulting in three different interpretations. #e conceptual spectrum agent-vehicle-patient, with the ensuing distinction between patient and vehicle is, as illustrated above in the examples with open and burn, by no means limited to descriptive verbs. For the semantic analysis of these particular verbs however, it is indispensable. Firstly, it admits a clear-cut barrier between the roles of agent and patient, and secondly it helps eliminate the confusion which would arise in explaining semantic distinctions if two such essentially differing participants as patient and vehicle were both referred to in the same terms. #is becomes clear for example, where verbs are used figuratively: 9a. $e sun glared down from the sky. (Vehicle) 9b. $e linguist glared at the pelican. (Agent) #is is one of several examples of the verbs denoting light-perception, which can also be used figuratively and agentively to describe facial expression, whereby the vehicular role of the sun is assumed by the eye. (Other examples are sparkle, gleam and glitter in English and blitzen, funkeln and leuchten in German.) #e modificant, centring round the dynamic adjectives fierce, hostile, intense, is in both senses of the verb glare basically the same; the constellation of roles, given the figurative variation, remains essentially stable. In conventional terms of case grammar however, the linguist (9b) would be classified as the Agent, while the sun (9a) would be the Patient; in other words, what is in fact functionally identical would be rendered in terms of binary opposites, in terms of a clear-cut dichotomy. a. Structure of roles in verb-descriptivity #e basic division of roles into participants and background circumstances, was illustrated above on the basis of strut; of these however, only the agent was relevant, the remaining participants and circumstances are in strut left open. #e semantic roles in descriptive verbs can however be much more complicated. Apart from the agent, vehicle and patient as discussed here, we have the adressee, the experiencer, the instrument and medium as participants, while the circumstances involve the surroundings, hierarchy of participants, cause, purpose and result of the verbal action. Among these roles, some tend to be more salient – and more complex – than others. 12 #e circumstances, as background, attendant factors in the verbal action, tend to be the least salient and the least complex. #e same applies to the medium (as for example the hampering medium such as snow in stapfen or dimmed visibility in peer) and the addressee (the participant to whom the action is addressed, as in grovel, schwindeln, glare and glotzen, but who, unlike the patient, can remain unaffected by it), insofar as they both give rise to and influence the action, but are only directly involved in its execution. #e remaining five roles are more salient and considerably more complex: while the conceptual essence of agent, vehicle and patient should be clear from the above discussion, in concrete realization there are frequent overlappings. It helps considerably if we take over the concepts of experiencer and instrument as auxiliary roles, admitting the possibility of role combinations. In verb-descriptivity the evaluating, normative part played by the speaker provides a further complicating factor, in that it colours and often even manipulates the roles of the different participants and their relation to each other. Let us illustrate this by comparing the English and German semantic fields of verbs denoting the expression of anger, as depicted in Snell-Hornby 1983: curse ----- swear fluchen ----- schimpfen storm scold wüten fume ----- rage rail toben bristle rasen geifern / keifen ---- zetern Here the experiencer of the emotion is the vehicle of the verb to the extent that he feels (salient in fume and bristle), and the agent to the extent that he acts (salient in curse and fluchen). Where an inanimate natural phenomenon is involved, such as a storm in rage or wüten, the verb is of course vehicular. #e subtle interplay of roles and the speaker’s manipulation of them is clearly revealed by a semantic analysis, both paradigmatic and contrastive, of the verbs concerned. Schimpfen for example focusses on the loud and voluble expression of anger and casts no aspirations on the agent, while fluchen focusses on the strong language used, but without indicating the cause, which must be specified in the context. #us fluchen covers curse and one sense of swear, whereby swear focusses on the actual verbs used and curse on the vehement imprecations behind them. Scold and rail focus on blame and fault-finding, whereby scold indicates noisy and voluble reproof directed personally at an addressee usually in a weaker position than the agent, typically a child or servant, and rail focusses on a furious outbreak of words aimed either at a person, whether present or absent, or at some object or abstraction. Storm and rage denote a state of violent fury, whereby storm, with a human agent, focusses either on angry movement or on expression through words, and rage, with either a human agent or an inanimate vehicle, focusses on the fit of anger or state of violence itself. #ese words are matched in German by wüten, toben and rasen, representing in that order a gradation of violence. Wüten focusses on an unrestrained and destructive outburst of aggression directed at some specific cause or object; toben denotes a fit of rage without a clear target or direction, but which so overcomes the agent that he loses all self-control; rasen is again directed at a particular object or cause, whereby the agent is driven into a state of frenzied fury. Fume and bristle both denote repressed anger, whereby bristle indicates a mute, static reaction of indignation and fume a state of seething anger or suppressed irritation, sometimes finding expression in words. German geifern denotes harsh and vituperative abuse, with extremely negative evaluation by the speaker and open condemnation of the agent. Keifen indicates venomous abuse made in a state of extreme agitation and in an unpleasantly loud shrill voice, hence invariably with a female agent towards whom the speaker expresses overt rejection. Zetern indicates a loud, persistent and quarrelsome type of fault-finding, whereby however the speaker detaches himself from the action and tends to find it amusing rather than repulsive. It is significant that what is objectively the same verbal action can sometimes be lexicalized by several of the verbs sketched above, whereby the difference lies in the attitude and perception of the speaker. (Cf. Snell-Hornby 1983, 115-6) A final comment must now be made on a problem arising with the distinction between vehicle and patient. Consider the following examples: 10 a) $e torch shone brightly. b) A torch was shone in my face. According to our interpretation, the torch in (10a) is the vehicle and a torch in (10b) is the patient. #at there is a fallacy here is shown in the German equivalents of these sentences: 10 c) Die Taschenlampe leuchtete hell. d) Jemand leuchtete mir mit der Taschenlampe ins Gesicht. #e surface realization of the German sentence discloses the instrumental character inherent in the role of a torch (10b), thereby revealing yet another idiosyncratic usage in English grammar. Here again, we may accept that, while the abstract concepts of patient and instrument are clearly distinguishable, in their concrete realization the two roles may merge and combine, according to the perspective of the sentence. In this way, our conceptual spectrum of agent – vehicle – patient is refined and completed; with the intermediary roles of experiencer and instrument supporting and combining with them, the entire continuum of salient roles in verb-decriptivity is covered, whereby we observe that the agent and experiencer are invariably animate, the instrument is inanimate, and the vehicle and patient may be either. As an offshoot of transformational generative grammar, case grammar naturally inherited some of its basic aims, methods and theoretical implications, among the most central of these being its syntactic approach, the dichotomy of deep and surface structure, and the aim of identifying language universals. None of these principles has been basic here: I rather hope to have pointed out the interdependence of syntax and semantics, of deep and surface structure and the the significance of idiosyncrasies specific to individual languages. It is interesting that the main problem of case grammar has been seen to lie in the delimitation and definition of the case-roles, aptly described by Dillon (1977, 68) as “intuitionism run wild”, by Pleines (1978, 360) as “Zaubertrick-Kasus”, and more soberly expressed by Fillmore himself: #e next truly worrisome criticism of case theory is the observation that nobody working within the various versions of grammars with ‘cases’ has come up with a principled way of defining the cases, or principled procedures for determining how many cases there are, or for determining when you are faced with two cases that happen to have something in common as opposed to one case that has two variants. (1977, 14) It was not my primary aim with this essay to enter the theoretical discussion on case grammar, nor did I intend with the introduction of the role of vehicle to aggravate the already existing “case proliferation”. #e roles discussed here enmerged from essentially empirical observation, and apply specifically to German and English descriptive verbs in their concrete realization. Our observations do however permit certain general conclusions, especially with regard to the initial concept of case roles as a limited set of separate and universally valid labels, whereby each role in its surface realization can be neatly inserted into its own theoretical slot. Later publications on case grammar disputed this view: Pleines, in his rejection of the claim to universality, described the “Zusammenspiel kognitiver, perzeptueller und linguistischer Determinanten” (1978, 368), while Van Valin and Foley emphasize the “interaction of syntax, semantics and pragmatics” (1980, 331), both arguments fully endorsed by this essay, where case roles have emerged as essentially pragmatic relationships within the semantic structure of the verb rather than as purely theoretical categories. One should distinguish between the purety of a theoretical concept and a fluidity apparent in its concrete realization. Dillon observes, with reference to semantic roles: … these concepts have a central core or prototype that most analysts would agree to and a number of associated properties, but they have fringes and overlappings with the fringes of other roles where a decision to assign one role instead of another will inevitably be somewhat arbitrary. (1977, 70) Accepting this principle of a central core with fringes and overlappings, we must also accept that no terminology is likely to present a fool-proof system of labels for every usage of every verb. It is unlikely that semantic roles will be conclusively arranged into a rigid system of slots and labels: in our observations they have rather emerged as a flexible structure of shifting, merging and interdependent components, reflecting, as Mario Wandruszka described it, the “schöpferische Unschärfe” of language.