Per Durst-Andersen* Towards a Multifunctional Grammar. ‘Language, Reality and Mind’ in a Grammatical Description

Previous grammars of the Russian language are written on the same methodological background and with the same purpose and may therefore be characterized in their entirety. It appears that they (1) are oriented towards the interpretive function, i.e. the hearer, (2) describe the different parts of the grammar in isolation without internal connection, (3) lack a contrastive element and finally (4) incorporate only written sources. In that respect previous grammars fail and cannot live up to what could be called modern standards. Against this background a new type of grammar is proposed a grammar which (1) takes the speaker into consideration, (2) differentiates three types of “wrongness”, (3) views the Russian language as a specific member of a linguistic supertype which is opposed to two other supertypes, and (4) takes its starting point in speech production, i.e. in oral discourse. After a theoretical discussion several pieces of evidence will be presented in favour of such type of grammar. 1. The traditional approach in Russian grammar Although new grammars are written in order to compensate for shortcomings experienced with previous grammars, they need not include a new framework or be based on totally new principles. In fact, it appears that relatively new grammars of the Russian language (e.g. Wade 1992, Mathiassen 1990, Christensen 1992, Švedova & Lopatin 1989) do not distinguish themselves in any way from previous grammars (Isačenko 1975, Mulisch 1975, TJAG 1979, AG 1980). It seems more or less as if the new grammars are echoes of their predecessors with an admixture of modern pedagogical principles, which need not be an improvement if simplicity supersedes insightfulness. In the following I 75 Hermes, Journal of Linguistics no. 17 1996 * Per Durst-Andersen Modern Languages. Russian Section Copenhagen Business School 15 Dalgas Have DK-2000 Frederiksberg E-mail: Durst/fra@CBS.DK shall attempt to describe what the abovementioned grammars have in common from the point of view methodology and current grammatical practice. Although I will confine my criticism to Russian grammars, I think that the four characteristics to be examined below are applicable to grammars of other languages as well. 1.1. The orientation towards interpretation A grammar may be used in two ways: to interpret or to produce spoken or written utterances. All existing grammars of the Russian language are constructed solely on an interpretive basis which means that they take their starting point in reception (comprehension). An interpretively oriented grammar takes some data as input, e.g. an utterance involving the perfective aspect, and yields a meaning as output, e.g. an action viewed in its totality. In other words, an interpretation can be said to have the form of an inductive inference: When testing what will be called the interpretive power of a grammar one could set up three requirements corresponding to the three boxes. The descriptions resulting from an interpretively adequate grammar should be (1) exhaustive, i.e. it should be suited to handle all data; (2) unequivocal, i.e. at best, the same rule or feature should be applied to all data or, at least, the reader should know which rules should be applied to which data; and (3) meaningful, i.e. it should give interpretations that are relevant and precise in meaning (translations to the readers’ mothertongue are not enough to satisfy this requirement). It is not my intention here to go into further details with respect to the three requirements I only want to point out that it is extremely difficult if not totally impossible for a grammar to satisfy the last two mentioned requirements if it is hearer-oriented. In order to satisfy requirements (2) and (3) we need a speaker-oriented grammatical component with a speaker-related notion as intention, i.e. the notion that corresponds to the pragmatic notion of sense (mening in Danish). 76 DATA GRAMMAR MEANING + ———> 1.2. The atomistic mode of presentation It goes without saying that any existing grammar has been written on certain principles which together can be said to form the leitmotif of the entire grammar. In our case it is quite evident that all are based on structuralist principles, where one of the most important principles says that language constitutes a structured whole in which everything is tied together. This leading principle of structuralism is not, however, reflected in traditional grammars. In other words, the verbal categories of tense, aspect, and mood might be described in the same structuralist feature framework, but you will not find the description that tells you how the three verbal categories form a structured whole and how they are tied together. In general, the different parts of the Russian language are described in isolation and without internal coherence. Not only verbal, but also nominal categories such as gender, animacy, number and case are described as if they had no relation to one another as if they did not exhibit any hierarchical relations whatsoever. The reader is not told neither explicitly nor implicitly by order of presentation whether aspect operates on tense or vice versa and which category is the basic one. In the same way the reader is not informed about the lexical-grammatical structure of verbs and nouns, and where lexical meaning could fit in a grammatical description. The order of presentation seems to be governed by the pedagogical principle of simplicity: what appears to be simple comes at first, what appears to be complicated comes at the end. In that way aspect, which is considered to be the most complicated category, may be treated as the final category (cf. Mathiassen 1990) an order of presentation which makes sense only from the point of view of the layman, not from the point of view of the Russian language. 1.3. The firm belief in Universal Grammar Modern linguists from the theories of Government and Binding across Functional Grammar to Cognitive Linguistics believe in what is normally called Universal Grammar (UG). According to this view all languages are grounded on the same basic structures and principles, are concrete manifestations of something invariant. The ultimate goal of linguistics is to account for UG and to describe and explain the various paths that can be taken by various languages. Although the three linguistic schools just mentioned have not yet created a total description of a single language from A to Z, the same underlying idea is found in 77


The traditional approach in Russian grammar
Although new grammars are written in order to compensate for shortcomings experienced with previous grammars, they need not include a new framework or be based on totally new principles. In fact, it appears that relatively new grammars of the Russian language (e.g. Wade 1992, Mathiassen 1990, Christensen 1992, Š vedova & Lopatin 1989 do not distinguish themselves in any way from previous grammars (Isačenko 1975, Mulisch 1975, TJAG 1979, AG 1980. It seems -more or lessas if the new grammars are echoes of their predecessors with an admixture of modern pedagogical principles, which need not be an improvement if simplicity supersedes insightfulness. In the following I shall attempt to describe what the above-mentioned grammars have in common from the point of view methodology and current grammatical practice. Although I will confine my criticism to Russian grammars, I think that the four characteristics to be examined below are applicable to grammars of other languages as well.

The orientation towards interpretation
A grammar may be used in two ways: to interpret or to produce spoken or written utterances. All existing grammars of the Russian language are constructed solely on an interpretive basis which means that they take their starting point in reception (comprehension). An interpretively oriented grammar takes some data as input, e.g. an utterance involving the perfective aspect, and yields a meaning as output, e.g. an action viewed in its totality. In other words, an interpretation can be said to have the form of an inductive inference: When testing what will be called the interpretive power of a grammar one could set up three requirements corresponding to the three boxes. The descriptions resulting from an interpretively adequate grammar should be (1) exhaustive, i.e. it should be suited to handle all data; (2) unequivocal, i.e. at best, the same rule or feature should be applied to all data or, at least, the reader should know which rules should be applied to which data; and (3) meaningful, i.e. it should give interpretations that are relevant and precise in meaning (translations to the readers' mothertongue are not enough to satisfy this requirement). It is not my intention here to go into further details with respect to the three requirements -I only want to point out that it is extremely difficult -if not totally impossible -for a grammar to satisfy the last two mentioned requirements if it is hearer-oriented. In order to satisfy requirements (2) and (3) we need a speaker-oriented grammatical component with a speaker-related notion as intention, i.e. the notion that corresponds to the pragmatic notion of sense (mening in Danish).

The atomistic mode of presentation
It goes without saying that any existing grammar has been written on certain principles which together can be said to form the leitmotif of the entire grammar. In our case it is quite evident that all are based on structuralist principles, where one of the most important principles says that language constitutes a structured whole in which everything is tied together. This leading principle of structuralism is not, however, reflected in traditional grammars. In other words, the verbal categories of tense, aspect, and mood might be described in the same structuralist feature framework, but you will not find the description that tells you how the three verbal categories form a structured whole and how they are tied together. In general, the different parts of the Russian language are described in isolation and without internal coherence. Not only verbal, but also nominal categories such as gender, animacy, number and case are described as if they had no relation to one another -as if they did not exhibit any hierarchical relations whatsoever. The reader is not told -neither explicitly nor implicitly by order of presentation -whether aspect operates on tense or vice versa and which category is the basic one. In the same way the reader is not informed about the lexical-grammatical structure of verbs and nouns, and where lexical meaning could fit in a grammatical description. The order of presentation seems to be governed by the pedagogical principle of simplicity: what appears to be simple comes at first, what appears to be complicated comes at the end. In that way aspect, which is considered to be the most complicated category, may be treated as the final category (cf. Mathiassen 1990) -an order of presentation which makes sense only from the point of view of the layman, not from the point of view of the Russian language.

The firm belief in Universal Grammar
Modern linguists from the theories of Government and Binding across Functional Grammar to Cognitive Linguistics believe in what is normally called Universal Grammar (UG). According to this view all languages are grounded on the same basic structures and principles, are concrete manifestations of something invariant. The ultimate goal of linguistics is to account for UG and to describe and explain the various paths that can be taken by various languages. Although the three linguistic schools just mentioned have not yet created a total description of a single language from A to Z, the same underlying idea is found in ordinary grammars. From the very descriptions and their conceptual apparatus it appears that grammarians take the Latin language (or the English language in a more extreme variant) as a kind of standard. The result is that the English and Russian tense-aspect-mood systems are described on the same basis. This leaves almost no room for differences between these two languages -contrastive analyses are kept at a minimum and often constrained to ad-hoc explanations. The heavy reliance on the old Latin tradition may also explain why only purely linguistic knowledge is taken as a possible framework of explanation. Neither extralinguistic knowledge (e.g. social rules) or interdisciplinary knowledge (e.g. psycholinguistic experiments) are taken into account.

The primacy of written sources
As already pointed out, traditional grammars are not only oriented towards the interpretive function, but also composed on the basis of samples of data taken from written sources. The data chosen, which are often picked out with great care from the very best authors of the last quarter of the 19th century, are necessarily grammatical, acceptable and appropriate. The one-sided, but quite understandable emphasis on wellknown and generally accepted material has several interrelated consequences. First of all, this type of grammar will not include examples from ordinary oral discourse -data which often differ in several respects from data from written discourse and which often would be more helpful for the student learning the foreign language. Secondly, this type of grammar operates -if it operates at all -with an extremely vaguely defined distinction between grammaticality and ungrammaticality -as if it was no concern of the grammar to explain why, when and in what way a sentence can be said to be grammatical or ungrammatical. This is of course necessary if the grammar is to reflect the grammar possessed by people who have the language as their first language.

The interpretive and predictive function of a grammar
There is a significant difference between a grammar that interprets utterances and a grammar that produces utterances. As just indicated, when a grammar interprets utterances from written discourse, they will be grammatical, acceptable and appropriate. When a grammar is to produce an utterance, it must adopt the speaker's position. In other words, it must account for the possibilities of selection and for the choice of grammatical forms in terms of the speaker's intentions, the knowledge derivable from the external situation or setting, and the background knowledge of the hearer, as well as the relationship between the speaker and the hearer. Thus, it does not necessarily follow that an interpretively adequate grammar is productively adequate, too. It is not necessarily capable of distinguishing grammatical and ungrammatical, acceptable and unacceptable, and appropriate and inappropriate outputs. (I here follow Hymes 1971 to a certain extent; for other views see Robin Lakoff 1977, van Dijk 1977, and Chomsky 1965 A productively oriented grammar takes a certain intention as input and yields an utterance as output. In other words, speech production takes the form of a deductive inference: A grammar which is interpretively as well as productively adequate should thus adopt the three way distinction presented above and should be capable of defining why and when an utterance or a text is ungrammatical (which has something to do with its external relation to reality as well as with its internal relation to other utterances), unacceptable (which has something to do with its relation to the hearer), or inappropriate (which has something to do with its relation to the speaker, i.e. his intention).

The holistic approach
It goes without saying that we need a psychological component in order to be able to talk about intentions and we need some tools to describe how intentions are formed and in which sort of language they are encoded. Since GB-theory sees no distinction whatsoever between linguistic grammar and mental grammar, i.e. between language and mind, and since Cognitive Linguistics identifies apperception (involving all senses) with perception (involving only the sense of vision), there is really no help to find in linguistics. I believe that there is a universal mental language which has its own grammar -the mental INTENTION GRAMMAR DATA + ---> grammar of the human mind. All other things being equal, people are borne with the same cognitive abilities which are evoked during childhood in much the same way regardless of speech community membership. The universal mental grammar is viewed as involving three different levels of representation. In other words, according to this view the same part of reality is represented in three different ways corresponding to the three different levels. It is thus the child's coupling of data with either of the three levels of mental representation that is responsible for the different types of grammar. Let me briefly examine what I consider to form the basic structure of the human mind. We shall take our point of departure in two different types of situation: (1) simple situations, i.e. states and activities, and (2) complex situations, i.e. events and processes. The two simple situations are identified on a perceptual basis -activities evoke unstable pictures, but states do notin opposition to the complex situations, viz. events and processes which cannot be identified on a perceptual basis. I define an event as a change of state caused by an activity and a process as an activity intending to cause a change of state. As we see, an event and a process are exactly made up by the two simple situations, i.e. by an activity and a state.

The first level: Identification
All people can identify an event, i.e. a state (q) caused by an activity (p), without having seen the activity itself. It is enough to have at your disposal two opposite pieces of information about the same thing/person (ground) in order to identify an event. If you know that there is nothing on your table (¬ q) and then suddenly see a vase with flowers standing on the very same table (q), then you will know that this cannot be true in the same world. Therefore you will conclude that the state situation "there is nothing on the table" (¬ q) is true of a past world, whereas the state situation "there is a vase with flowers on the table" (q) is true of a present world. In fact you have identified a change of state caused by some activity since you know that no change of state (i.e. from ¬ q to q) can take place without it having been caused by some activity (cf. von Wright 1974). This is all done automatically and subconsciously on the basis of what I call the event-instruction (see fig.1). An event is thus identified and experienced indirectly without you seeing the activity itself. You may also identify a process, i.e. an activity (p) intending to cause a new state (q). A process is experienced forwards, since you must see an activity in order to be able to infer that there is a certain intention underlying the production of that activity. This is done on the basis of receiving an unstable picture on your perceptual screen. Let us imagine that you see your mother carrying a vase with flowers. You identify an activity by receiving an unstable picture. But this is not enough to identify a process. You have to be able to see that your mother is walking towards a certain table and conclude from this background that she has the intention of putting the flowers on the table. This need not turn out to be true in view of the fact that she may turn left and put them on the window sill instead. But if it turns out to be true then the identified proces becomes an event which is directly experienced. This means that the event instruction gives a correct output in all cases, whereas the process instruction does not garantee a correct output: it involves an abductive element, i.e. pure guessing.

The second level: Assimilation
Now the identified event or process has to be assimilated, i.e. interiorized into the human mind. The assimilation is provided by the mental models of events and processes (cf. fig.1). Such mental models, which are generalized, abstract copies of concrete events and processes of reality, must exist because without them we could not account for the fact that in the case of an identified event an unobserved past activity suddenly becomes part of our human mind and in the case of an identified process it is an unobservable future state that suddenly becomes part of our human mind in the shape of an imagined state. We cannot possibly explain this without assuming that people have at their disposal generalized copies of these complex situations, i.e. mental models of events and processes. In other words, at the level of mental models identified events and processes -of which you have actually seen only one of two parts -are internalized into the human mind as complete structures.

The third level: Storing
After having assimilated an identified event or an identified process people must store them in order to be able to assimilate new events and processes and in order to be able to recall or visualize them, and to talk about them later on. From the point of view of normal human abilities at least two different kinds of mental stores must be available to the human mind: one of the past world and another of the present world. This must be the case since people are capable of differentiating past experiences which have no present effects and those which have. For instance, If we stick to the flower-scene and presume that you have experienced the event indirectly, i.e. identified it backwards, you know that the state situation "there is a vase with flowers on the table" obtains in the present world, whereas the activity that lead to the present state does not obtain anymore, i.e. it is, and will always be in the past. The event can and will only be stored as a photograph in the present world store -you have not seen the activity which would have left a film in the past world store. If the event was directly experienced, i.e. identified as a process which later turns into an event, then you will have it stored as a film in the past world store -you actually saw the activity component -but also as a photograph in the present world store, because the state situation "there is a vase with flowers on the table" still obtains. When the vase is removed, the flowers thrown out or the vase is broken, the photograph in question will be "burnt": the original situation is gone and therefore the copy taken from the original has to be removed from the present world store to the past world store.
The important and interesting thing about this is that the three levels reflect three different mental representations of the very same event and process: (1) at the level of instructions people identify events or processes either directly or indirectly; (2) at the level of mental models people assimilate or interiorize the identified event and process by using the mental models of events and processes, respectively; and (3) at the level of stores people store the assimilated events and processes in the present and past world stores, respectively. It it my hypothesis that languages may choose to verbalize either of these three levels of mental representation. In other words, a language may choose to verbalize what has been directly and indirectly experienced, or to verbalize the mental models which are generalized copies of reality or to verbalize the past and present world stores. Whatever a language chooses to do, the grammatical outcome will be quite distinct from the other two possibilities.

Linguistic functions and linguistic supertypes
If we compare these three established levels of the human mind to Bühler's Organon Model (cf. Bühler 1933), it appears that there is a clear connection (see fig. 2). Bühler operates only with three so-called language functions in opposition to Jakobson and Halliday who operated with up to seven different functions thus -as I see it -confusing the notion of language function with the notion of speech function. Each function is attached to one of the three obligatory participants of any communication situation, viz. the speaker who has a message, the hearer for whom the information is intended and objects and situations of reality which can be said to be carried by the linguistic notions of words and sentences. The expressive function is thus defined in relation to the speaker, the appeal function in relation to the hearer and the representative function in relation to external reality. The three language functions correspond not only to the three traditionally recognized sign types, viz. symptom, signal and symbol (cf. fig.2), but also to the hierarchical structure of the human mind just examined above (cf. fig.1). Thus, on the basis of Bühler's Organon Model (1933) and on the basis of the three levels of representation (for further information, see Durst-Andersen 1992: 81-96) I suggest that the grammatical categories of any single language and their formal coding are designed to capture primarily one of the following: (1) the speaker's encoding of message, (2) the information which is intended for the hearer and which is decoded by the hearer, or (3) situations in referential reality which are represented in the hearer as well as in the speaker. This yields three linguistic supertypes: (1) speaker-based languages like Georgian and Turkish (level one), (2) hearer-based languages like English and Danish (level three), and (3) reality-based languages like Russian and the informal version of French (level two). Each such supertype has its own core category -speaker-based languages have a mood distinction between direct and indirect experience, hearer-based languages have a tense distinction between flash-back (old information) and news-flash (new information), and reality-based languages have an aspectual distinction between events and processes (see fig.3). Each core category which determines the entire core system, thereby creating internal harmony among its members. The internal harmony is not only argued to be a fact inside other verbal categories, but also inside nominal categories because the nominal system appears to be isomorphic with respect to the verbal one (see below). If this holds true, then we have an explanation of the fact that when looking at the grammatical categories of the Russian and the English languages it is not possible to find a single common category -it is either the case that a category is present in one language (e.g. definite and indefinite article in English or case in Russian), but absent in the other, or the case that a category in one language (e.g. aspect, tense, and mood in English) is completely different from that of the other language -i.e. what seems to be the same because they have the same name cannot be put under the same heading and therefore do not fit into the same structure (see also Durst-Andersen 1993/94).

Supertypes and form vs. substance
Although the notion of linguistic supertype is new, it appears that the idea of making a distinction between type (supertype in my terminology) and system is not new at all. In 1965 Coseriu proposed a tetratomy of language, i.e. a four-way distinction, which included type, system, norm, and speech (cf. Coseriu 1965). By so doing, Coseriu split the Saussurian notion langue and the Hjelmslevian notion form into three 85 MOOD

Direct experience of events
Koydu (-dI)

Indirect experience of events Koymuş (-muš)
Action as a process Stavil (ipf) Action as an event Postavil (pf)

Event as a flash-back Put (simple past)
Event as a news-flash Has put (pres.perf.)

HEARER-BASED LANGUAGES
parts: (1) type which has something to do with principles of organization, (2) system which is the materialization of some more or less abstract principles into rules, and (3) norm which is the normal realization of the rules. This means that some principles can be organised or weighed differently and the result will be different systems, and in the same way a certain system, which is in fact something potential, can manifest itself in many ways, but one of them is the normal one. Let us say that is possible to distinguish between consonantal and vocalic languages (cf. Andersen 1978), i.e. between two different types. Russian and Polish would be consonantal languages because they both have the largest number of phonological distinctions inside the consonants. Their phonological systems are, however, different, since the interpretation of the principles have resulted in different rules. The Russian system has several manifestations -i-kan'e vs. je-kan'e, a-kan'e vs. okan'e, etc. -where it is possible to find the normal realization (which is i-kan'e and a-kan'e). The norm can only be found by looking at the substance, i.e. speech in Coseriu's terminology. In other words, at the substance level Saussure's notion parole equals Coseriu's speech.
From this the conclusion to be drawn is that although two languages belong to the same linguistic supertype (or type in Coseriu's terminology), they may manifest themselves in two different systems. In the former case we are dealing with identical principles, in the latter case we are dealing with different interpretations which result in different rules. By looking at a particular system of a specific linguistic supertype (e.g. Russian as a member of reality-based languages) and by comparing it with a specific system of another linguistic supertype (e.g. English as a member of hearer-based languages), we automatically get a contrastive element. In other words, the proposed theory of linguistic supertypes is a holistic typology which ensures that it is the entire system of L 2 that is compared to the entire system of L 1 . This is crucial, since only in that way it will be possible to convey the "biological nature" of the language to foreign students. (For a "total-grammatical" approach, see Hawkins 1986 andMüller-Gotoma 1992.)

Word-for-word vs. sentence-for-sentence translation
Let us take an illustrative example and let us test how this kind of grammar operates. Let us imagine that some Danish students have been assigned the task of translating the short story "Buketten" by Leck Fischer into Russian. The first sentence to be translated is the following: (1) Der blev banket på døren 'There was a knocking at the door.' there-adv knock-pas-pret on door-def From my own experience as both a student and a teacher of Russian I know that the Danish student will not have the faintest idea of what to do with that sentence: there is no grammar and there is no dictionary that might be helpful. He will probably start by taking the dictionary and try to look under the entry der 'there'; make some inquiries about passives in a Russian grammar; once again look into the dictionary to get some help to translate the preposition på 'on'; and finally consult the grammar again to find the exact case form to be used in the given utterance. This approach where the student analyses and translates in components, i.e. word for word, is completely wrong from the point of view of the Russian language -one might argue that it actually jumps over the Russian grammar. As a reality-based language, Russian is designed to capture situations in reality. In view of the fact that the Danish language is designed to capture the information intended for the hearer, the student has to imagine which situation he thinks is referred to by (1), i.e. he has to choose one of several interpretations corresponding to different situations. The Danish student would be able to arrive at the "relevant" interpretation by deciding on the following variables: -Type of situation (state, activity or action)? → activity verb (stučat') -Past or present activity? → past activity (-l) -Presentation of the activity (act, process or event)? → event (pf) -Duration of activity itself (short, too long, without interruption)? → short (po-) -Individuation of agent (agent specification or not)? → none (-i) -Is the activity vertically or horizontrally directed? → vertically (v) -Is it permanent or is sporadic contact? → sporadic (acc) All these question should be asked and answered in order to arrive at the Russian utterance which can be said to match the most likely interpretation of the Danish utterance in the given context. In that way we get (1').
(1') Po-stuča-l-i v dver'-Ø pf-knock/ACT-pret-plur in door-acc The Russian utterance has the following semantics: "Somebody who is not important to our story went from producing two or three knocks at a vertically placed door to not producing knocks". As should be evident, although in principle the Russian utterance talks about the same situation as the original Danish utterance, they cannot be said to be a hundred percent identical in meaning -on the contrary. The Danish language acts as an appeal to the hearer/reader and is designed to capture the distinction between new and old information at various levels. Der 'there' signals a so-called presentative sentence, i.e. introduces a new frame of reference (a new story or a new part of an old story), the simple past form shows that this part of the story is presented as a flash-back, and the blive-passive that it is shown as a film (not presented as information bites). Note that the Danish utterance says nothing about the activity component. It could be a single act, two or three knocks or in principle an ongoing knocking activity. Only the reader or the hearer can decide what it looked like. In other words, external reality is treated as something that the reader has to construct for himself -in direct contrast to the Russian language.

Translation and interdisciplinary knowledge
If we go a little bit further down the text it turns out that some person had come to deliver some flowers to Mrs. Elly. These flowers are however received by her husband. At the point where he shouts to her that the flowers are for her, we realize that it is not sufficient for the student to have the "valid" grammar, he must also possess knowledge about social rules of the Russian society and be capable of applying textlinguistic notions: (2) "De er til dig". Hun afbrød ikke arbejdet 'They are for you! -She (Mrs.Elly) didn't stop her work.' The question is how to translate the negated utterance. Normally it is recommended by Russian grammars that the imperfective aspect should be used when the potential agent did not do anything. I made an experiment with Russian native speakers who master the Danish language in which I asked them to translate the negated utterance. First I did not show them the continuation of the text. All chose the perfective aspect and said that the imperfective sentence would be ungrammatical: (3) On-a ne prerva-l-a svoj-u rabot-u she-nom neg stop/pf-pret-fem her-acc work-acc Then I showed them the entire text where it is stated that Mrs. Elly did not stop her work because she did not want to receive the flowers -her husband had sent them in order to calm her down (he had been fooling around with another woman). At that point they all changed opinion and said that the perfective sentence was ungrammatical. Why did they change their opinion? In the first case they chose on the basis of their knowledge of social rules. When women are brought flowers, they are supposed to behave according to the rule saying that they should stop whatever they are doing in order to be able to receive the flowers -the opposite would be impolite. In other words, it was presupposed that Mrs. Elly was obligated to stop her work. This specific type of presupposition can be conveyed only by the perfective aspect. When the Russian informants realized that she in fact did not want to receive the flowers, they changed opinion because only the imperfective aspect implies that the potential agent did not want to do anything (see 3a).
A normal grammar which does not operate with social rules, expectations and textlinguistic notions would not be able to handle this kind of data.

Semantically different, but pragmatically identical
Although languages belong to different supertypes, it does not necessarily follow that it is totally impossible to bring differently constructed grammars under the same heading. Let us take an illustative example. In Russian there are a perfective as well as an imperfective imperative which are used to issue MANDS at the expence of indirect speech acts, which are used far more often in Danish than in Russian, but in Danish less frequently than in English, where in most cases direct speech acts would imply something like a military command (e.g. Sit!, Go!, Get up!, etc.). These differences have something to do with the fact that Russian, on the one hand, and English and Danish on the other, are grounded on a completely different semantic basis (they belong to different supertypes), and with the fact that English and Danish are different manifestations of the same supertype, i.e. have different rules although founded on the same principles (for a cultural explanation, see Wierzbicka 1991). However, by looking at various parameters at a higher level it is possible to find correlations and make them fit into the same pragmatic framework. In that way it turns out that it is possible to describe (a) the use of aspect in connection with the Russian imperative mood, (b) the use of Danish "modalizers" (i.e. modal particles) in connection with imperative sentences, and (c) the use of indirect speech acts in English (see fig.4). From the pragmatically defined conditions inside what is here called "The Instructions for Directives" it appears that the major condition is whether the speaker considers the hearer to be cooperative or not. If he decides YES (either because the hearer desires the same state as the speaker does or because he wants to be helpful with regard to the speaker's desire), the speaker must decide whether it is true or false that the hearer dares to produce the required activity. If he decides YES, he will use the perfective aspect in Russian, the LIGE-particle in Danish, and "Will you...?" in English. If he decides NO, then he will use the imperfective aspect in Russian, the BARE-particle in Danish, and "Won't you...?" in English. If the speaker decides that the hearer is not cooperative (because the hearer hesitates to perform the requested activity), he must decide whether it is true or false that the hearer desires the same state as he does. If he decides YES, he will use the perfective aspect with the že-particle in Russian, the NU-particle in Danish, and "Can't you...!" in English. If he decides NO (i.e. the hearer hesitates to produce the activity, simply because he does not desire the state which will obtain if he produces the activity), he will use the imperfective aspect with the the že-particle in Russian, the SÅ-particle in Danish, and "Will you...!" in English. It appears that Russian distinguishes between deontic and alethic modality by means of aspect and between polite and impolite uses by means of particle and/or intonation, that Danish distinguishes between the various deontic and alethic modes by means of different particles ("modalizers"), and English distinguishes between polite and impolite uses by means of "sentence mode" -interrogative sentences sig-nalling politeness, whereas declarative sentences with inversion of subject-predicate signal impoliteness.

Fig.4: Instructions for Directives
Although different social rules may interfere and cause some local mismatches in the above-mentioned correlations, it ought to be clear that we have some simple tools to make foreign speakers learn to produce grammatical utterances which are simultaneously appropriate from the point of view of the speaker, as well as acceptable from the point of view of the hearer.

Scenario descriptions and oral discourse
Another way of making the grammar more suited to satisfy the predictive function of a grammar is to use scenario descriptions where the student/reader has to choose the utterance that fits the description: Scenario description 1: The hospital scene An old lady is standing in front of a couple of chairs in a hospital hall and is not sure if she may sit down as she wants. You, who are a doctor, passes her and observes her uncertainty. In order to make her sit down you say: If we assume that the task was given to a student who chose the right one, viz. (4d), he might have explained it in the following way. In the situation described there is really no choice in Danish, simply because all examples from (a) to (c) will imply that they are uttered in the doctor's own interest, which is wrong according to the scenario description. Only (d) makes sense: the old lady needs a permission from a person who has the right authority and (d) with bare involves the necessary licence and is uttered in the interest of the old lady. Bare 'merely' is thus a prescriptive modalizer, the function of which is to issue a permission and by doing that to satisfy the hearer's precondition for complying with the speaker's request. Such scenario descriptions will not only help the student to produce speech, but at the same time ensure that we do not forget oral communication and oral language (which is so poorly represented in ordinary grammars).

The grammar itself -evidence from Russian
One of the advantages of relying on the notion of linguistic supertype is that we can explain why, for instance, the Russian and English languages do not have a single category in common. Russian has aspect, but it differs fundamentally from English aspect (the difference be-tween the perfective and imperfective aspect is learnt by the Russian child in the past tense, whereas the difference between the progressive and non-progressive aspect is learnt by the English child in the present); Russian has tense, but it differs fundamentally from English tense in the sense that Russian tense talks about past or present situations, while English tense talks about past and present worlds, i.e. people/things are located in two different kinds of universe; etc. In addition to that, Russian has a whole arsenal of formalized categories which are simply absent in English (e.g. case, animacy, mood, etc.) just as the English language has (e.g. definiteness, it-and there-sentences, it-and therecleft sentences, etc.). In the same way, it becomes clear why it is difficult -if not impossible -for Russian students to learn to speak and write correct English as well as to comprehend oral and written English discourse, and, conversely, for English students to learn to speak and write correct Russian as well as to comprehend oral and written discourse. By saying -as we did above -that there is a determining category, i.e. a category that governs the entire system, we automatically get the idea that some categories are more important than others. This is, of course, extremely important for acquiring a second language and should be reflected in any grammar. This brings us to the question of the hierarchical organization of verbal and nominal categories and to the question of isomorphism between the structure of the verbal and the nominal systems.

The verbal categories and their internal structure
In 2.1. we talked about state (q-classifiers), activity (p-classifiers) and action verbs (p-and-q-classifiers). Normally they are called verb classes or aktionsarten outside the Russian tradition. Without going into further detail I shall say that these three classes comprise an important grammatical category which has been overlooked so far. The function of this verbal category can be said to be equivalent to gender and for that reason I shall call it verbal gender (VG). In this way, the Russian language can be said to have three verbal genders -state verbs which create a single ground-proposition involving a state description, activity verbs which create a single ground-proposition involving an activity description, and action verbs which create two ground-propositionsone involving an activity description and another involving a state description. Due to the fact that state and activity verbs involve single ground-propositions, but action verbs two ground-propositions the former can be termed simplex verbs and the latter complex verbs.
What is then the function of verbal gender? Its function is best understood in metaphorical terms. It can be said to create an IMAGE-IDEA PAIR, i.e. something potential which has not yet been actualized and reached the real world. For instance, both the perfective and imperfective action verb dat'(pf)/davat'(ipf) 'give' create two ideas in the form of two ground-propositions, viz. X DO SOMETHING (p) and Y EXIST WITH Z (q), both being coupled to two corresponding imagesone being unstable where X is doing something with Y and another being stable where Y is with Z. In that way we get a complex IMAGE-IDEA PAIR or a verb model of actions. The common meaning of the two aspectual forms in Russian is therefore not to be found in an autonomous component like the English verb "to give", but in the common root //da//-. Aspect (A) operates on the output structure of the category of verbal gender, and by doing that it can be said to create a copy of reality. An action as such is not found in reality: it is not possible to find a world where the activity of the action "X BRING Y TO Z" obtains at the same time as does the state of the action -either it must be the case that X did something that caused Y to be with Z (an event), or it must be the case that X is doing something in order to cause Y to be with Z (a process). This means that an action is a construct or a metaconcept of events and processes -just as it is not possible to find a human being as such: it will either be a male or a female variant. In other words, aspect brings reality into the picture and therefore we should no longer speak about X and Y, but about concrete human beings as Grigorij and things as a certain book. The function of the perfective aspect is to show to the hearer a concrete copy of an event, whereas the function of the imperfective aspect is to show a copy of a process (an action cannot create a copy at all because there is no original in our external reality). The category of tense (T) operates on the output structure of aspect (copy) and by doing so places the original in reality, i.e. either before or simultaneously with the moment of speech. The category of mood (M) operates on the output structure of tense (original) and by doing so assigns the relation of equality or non-equality between the copy and the original. Indicative assigns the relation of equality and thus indicates that we are dealing with true reference, i.e according to the speaker there is or was such a situation in reality. The two non-indicative forms assign the relation of non-equality. The subjunctive mood treats the lack of equality as a matter of fact, i.e. it cannot be changed. The imperative mood treats it as matter to be changed, i.e. the hearer is asked by the speaker to create an original on the basis of the copy (see fig.5).
This is the best way to explain the hierarchical structure and internal functioning of the system of verbal categories in Russian. It shows that verbal gender is a lexico-grammatical category, that aspect is a propositional-semantic category, that tense is a referential-semantic category and that mood is a relational-semantic category. This hierarchical structure is iconically reflected in the segmentational structure of morphemes: (5) give-pf-pret-subj da -Ø -l -by "(he) should have given (it)'

The nominal categories and their internal structure
The point is that the structure of the nominal system is isomorphic with respect to the structure of the verbal system -the former is a diagrammatization of the latter. This means that the four verbal categories mentioned above have their exact equivalents in the nominal system. Thus it seems to be no coincidence that the Russian language consists of four verbal categories as well as four nominal categories, viz. gender (NG), animacy (A), number (N) and case (C). A look at the order of the involved nominal morphemes should provide us with an idea of the hierachical structure on the level of content: (6) bror -pl-gen brat -j -ov G/A -N -C Just as verbal gender and verbal aspect, nominal gender and animacy inherently belong to the stem, whereas the remaining two categories belong to morphological endings. Nominal gender also creates an IMAGE-IDEA PAIR, i.e. in the case of e.g. myš 'mouse' we will have 95 COPY RELATION ORIGINAL ASPECT MOOD TENSE a prototypical description of a "mouse" paired to a protypical picture of a "mouse". Nominal gender is thus a lexico-grammatical category.
The function of the category of animacy can be compared to the function of verbal aspect, since it operates on the output structure of gender and by so doing shows the hearer a concrete copy of a "mouse" -in the case of an animate mouse the hearer is shown a normal living mouse (see 7a), in the case of an inanimate mouse (see 7b) the hearer is shown a computer mouse.
(7a) ja uvidel dvux myšej 'I saw two (genuine) mice'. I see-pf-pret two-gen-pl mouse-gen-pl (7b) ja uvidel dve myši 'I saw two (computer) mice'. I see-pf-pret two-acc-pl mouse-acc-pl The former will create a potentially unstable copy ("agentiveness"), the latter will create a stable copy ("passiveness"). In other words, the category of animacy/inanimacy could be termed an ideational-semantic category. The category of number operates on the output structure of the category of animacy and by so doing it places the original (for instance, two mice) in reality. In that respect its function is comparable to that of tense and can be named a referential-semantic category.
Case is equivalent to mood and is thus a relational-semantic category in the sense that it shows the hearer whether there is a relation of equality or not between the copy showed by the category of animacy and the original placed in reality by the category of number (see fig.6).
Direct cases, i.e. the nominative and accusative cases, function as the direct mood form, viz. the indicative form. They assign the relation of equality and thus signal that the person or thing denoted by the noun is indeed present in the situation described by the verb (I call this local reference). All other cases are oblique and assign lack of equality. The vocative case, which is included in the Russian case system because of its productivity (cf. Bily!1990), functions as the imperative mood: it treats the lack of equality as matter to be changed, whereas the remaining three cases (the genitive, dative and instrumental cases) treat the lack of equality as a matter of fact. The difference between them 96 COPY RELATION ORIGINAL ANIMACY CASE NUMBER seems to correspond to the distinction between objective mood (the genitive) and subjective mood (the dative and the instrumental). The grammar just described is capable of explaining what is inexplicable in ordinary grammars (if mentioned at all): in Russian we have to use the subjunctive mood in connection with the genitive case in the following utterance (the indicative mood will be ungrammatical): (8) Na Ukraine net ni odnogo goroda (gen), kotoryj ne golodal by (subjunctive) vo vremja vojny 'There is no town (gen) that did not suffer (subjunctive) from hunger during World War Two.' We can explain this phenomenon as a kind of harmony: because there is no local reference the genitive case is obligatory and by opening up an internal world of imagination, the speaker must continue to speak in the same terms and therefore has to choose the subjunctive mood which also assigns the relation of non-equality between the copy and the original. The distinction between local and non-local reference inside the Russian case system is also found in the pronominal system where pronouns with -to and ničego assign local reference, whereas pronouns with -nibud' and nečego assign non-local reference.

Lexicon -grammar -communication: nom. vs. instr.
As already pointed out, ordinary grammars only seldomly mention lexical meaning. This is a pity, since a certain grammatical way of thinking must reveal itself at all levels of language -also at the level of lexicon. Let me illustrate this by using the category of animacy vs. inanimacy, which plays a far more important role in the Russian language than normally acknowledged by grammarians. The grammatical distinction is instantiated in the lexicon as two different words for the English word corpse, viz. mertvec 'animate' and trup 'inanimate'. At an even higher level of abstraction the grammatical and lexical distinction can be explained in religious terms: according to orthodox religious thinking the soul leaves the human body only after 40 days. This creates the need for two different kinds of the notion "corpse".
Owing to the exclusion of lexical meaning, traditional grammars actually have nothing substantial to say about the combination of lexical and grammatical meaning. This phenomenon seems to be rather important, since, for instance, (9a) is grammatical, while (9b) is ungrammatical.
categorized as a poet and since this is known to every Russian person, it is felt as being impolite to say such a thing -as if the speaker could not presuppose such knowledge. In this specific case you would normally use the nominative case and in that way bring some new information into the utterance, viz. he was the poet per se. The distinction between the nominative and instrumental case is also found inside adjectives. The semantics is the same, i.e. the nominative case characterizes a person or a thing, whereas the instrumental case categorizes. However, when the adjective functions as a predicative determiner, i.e. when it occurs in a secondary nexus, the distinction in case marking becomes a text linguistic distinction between foregrounding and backgrounding (see 12 and 13).
(12) On-a vyš-l-a iz vod-y mokr-aja 'She went wet out of the water'. hum-F go-pret-F out water-gen wet-nom (13) On-a vyš-l-a iz vod-y mokr-oj 'She went wet out of the water'. hum-F go-pret-F out water-gen wet-instr The nominative foregrounds the state situation where the female person is described as being wet, while the instrumental case backgrounds itthis means that the state situation where she is out of the water is foregrounded instead (see also Timberlake 1986). As we can imagine, foregrounding and backgrounding necessarily implies a certain kind of harmony between various parts of information given by the preceding or the following utterance. For instance, it is completely coherent to use the nominative case in (14), but not completely harmonic to use the instrumental.
therefore later could take home). The backgrounding function of the instrumental case is evident in (15).
(15) Ona vyšla iz vody mokroj instr (mokraja nom ) i prostudi-l-as' 'She went wet out of the water and cought a cold' In (15) the nominative case would be out of the question -it is inappropriate from the point of view of the speaker's message -since it would focus on the information that she is wet thus requiring the focus to be continued in the following discourse (in the shape of a description of her looks). In contrast to this stands the instrumental case which backgrounds the information -it is appropriate in the given context. In that way the state where she is wet can function as the reason why she caught a cold, and the entire utterance thus gives the right associations to the hearer (cf. Karaulov 1993). The same phenomenon is found in aspectual usage and is documented in an interesting experiment with 100 Russian informants (see Svedova 1984). They were all given one hundred utterances including a certain context (either in the shape of a preceding utterance or in the shape of a following one) and were asked to choose aspect, i.e. the perfective or imperfective aspect. Let us take the following utterance as an illustrative example (see 16). It turned out that 97 pct of the informants marked the perfective aspect ungrammatical, while 3 pct of them found it was grammatical and even more preferable than the corresponding imperfective form. Both forms are in fact grammatical. This is a question of foregrounding vs. backgrounding of information or once again an instantiation of the phenomenon "communication-overrules-grammar". The perfective aspect foregrounds the state "The translation exists on world-location" and backgrounds the activity "Somebody produced an activity with that intention". Since the passive sentence asserts that the translation is clumsy and thereby presupposes the physical existence of the translation itself, the result is that there is mismatch in the information packaging. If something appears to be presupposed in a preceding affirmative utterance and against this background is asserted (i.e. foregrounded) in the following question, we get a tautological statement inside the question and the perfective aspect is judged ill-sounding -we feel that we are not getting further, but have stepped back. There is too much harmony between the two utterances. The imperfective utterance shifts topic, i.e. it introduces the question of a translator by foregrounding the activity description and in that way introduces something new so that the hearer will feel that we in a way are getting further.

Concluding remarks
I have just attempted to outline the fundamentals of a new type of grammar, called Mental Grammar, which operates with a sharp distinction between language, mind, and reality. It rejects the existence of UG by splitting it up into three different linguistic supertypes. On the background of various unexplained data from the Russian language, which is a so-called reality-based language, I tried to demonstrate that such type of grammar lives more up to what was called modern standards than previous grammars which are oriented only towards the hearer, rely almost totally on written sources, and lack a contrastive element. fi