Now one of the problems with the term word order is that we normally don't mean word order, but constituent ordering. My hope is that Clark's work will help us to explain not only constituent orering but individual word order as well, so I shall keep the term word order for the moment.
Another problem is that, even though we know a great deal about word order in German, and the ideas of functional sentence perspective offer the possibility of explaining the architecture of sentences, no over-arching rationale of both fixed and moveable word orders has been proposed. My hope is that Clark's work will allow us to do just that, and also propose a general explanation for word order in all phrasal types. In other words, I think we can use the same principles to describe word order in noun phrases and prepositional phrases as well as verb phrases.
"People use language for doing things with each other, and their use of language is itself a joint action." (Clark 1996:387).
Essentially, this is derived from Austin's How to do things with words (1962) approach, but the idea of joint actions is fronted more than in Austin's work and its derivatives. Clark goes further:
"Language is rarely an end in itself. It is primarily an instrument for carrying out broader activities [...] in which two or more people, in socially defined roles, carry out individual actions as parts of larger enterprises. Language is simply a device by which they coordinate those individual actions." (Clark 1996:387)
In other words, language is not autonomous and cannot be studied "without studying joint activities and vice versa" (Clark 1996:387). The close link between language use and joint activities has led people to see certain phenomena as features of language use when in fact they are features of the joint activities in which the language is being used. "These phenomena include coordination, cooperation, conventions, turns, closure, joint projects, opportunistic actions and the accumulation of common ground" (Clark 1996:388).
So I'm concentrating today on two aspect of these joint activities, first that of common ground, and second on the phenomenon of closure, as it appears to be an important factor in word order.
First, joint activities and the joint actions through which they proceed are impossible without an accumulation of common ground between the participants. Clark (1996:120-121) explains common ground as a form of self awareness, which is reflexive. In other words, participants must not only be aware of the information they have in common, that awareness itself must form part of the information complex. Common ground can be justified in terms of a shared basis in experience or action, but it is of variable quality. Some shared bases allow us to draw conclusions about people, but not others. Clark gives the example of an American adult, of whom one can be certain that they know the name of the current US president, but not necessarily the Thirteenth Amendment to the Constitution. He divides common ground into two broad types: communal common ground essentially derived from cultural knowledge; and personal common ground derived from personal acquaintance. Clark's work is based on spoken language, but I believe that his conclusions are applicable to written language as well, for example in the construction of text types according to norms accepted by text originators and text readers (see West 1990 for a discussion). This is after all a type of cultural knowledge. Indeed, language itself is a cultural phenomenon, so one could view the patterns observable in natural languages - from paradigms and lexical entries to sentence patterns and textual conventions - as a reflex of cultural common ground. If contributors and respondents did not share this information, communication would be impossible.
The idea of common ground scores over quasi-mathematical notions such as presupposition (see Allwood, Anderson and Dahl 1977 for a survey) because language is thereby freed from the straitjacket of a non-linguistic metalanguage. Indeed, common ground accords with what Lakoff (1987:xii-xv) calls an experiential rather than an objectivist view, which implies, among other things, that thought is not abstract, computational, disembodied, atomistic or logical, but is by contrast embodied, imaginative, has gestalt properties, what he calls an ecological structure and can be described using cognitive models which have these properties. Common ground helps us to see how people understand each other in the real world. Participants work hard to ground what they do together, in Clark's words (1996:221) to "establish it as part of common ground well enough for current purposes". To do this, Clark argues, participants also need closure on their actions, by which is meant the general principle (due to Norman 1988) that "agents performing an action require evidence, sufficient for current purposes, that they have succeeded in performing it" (1996:222). Participants in joint actions also look for evidence that they have achieved what they set out to do. "Contributors present signals to respondents, and then contributors and respondents work together to reach the mutual belief that the signals have been understood well enough for current purposes" (1996:252).
Clark (1996) applies these ideas to face-to-face interaction, and implies that action ladders are essentially cotemporal. In reality, of course, it takes time to perform even the simplest of actions - probably as much as half a second (McCrone 1999:124-131). If joint closure is required at each level of the simplest action ladder, this implies a processing time of at least two seconds. It may be that socio-cultural convention and cues from previous discourse result in a shorter "thinking time" than this, but we are nevertheless faced with the fact that even seemingly instantaneous processes are linear in nature. Indeed, this must be the way that longer stretches of discourse, and the accumulation of common ground which they entail, work. Once closure is achieved on a part of a joint activity, the information it implies is added to the information which is already part of common ground. It follows that common ground is laid down in a linear fashion and is cumulative in nature. If this deduction is valid, we should find that, within a given structure, the elements which are placed earlier are more closely rooted in common ground than those which are placed later. Furthermore, we should also find that, where there is a choice in sequencing, the alternative chosen by the speaker or writer will most likely be justifiable in terms of common ground. Of course, because ordering in this latter case is something that speakers and writers do in German, it is hard to set up hard and fast rules. By and large, it should be possible to predict that writers strive to make their texts consonant with the principle, and that speakers do this too, but that speakers have a chance to repair their sequencing if they find that it does not work in a given context. Writers have no such opportunity and this may account for the more rigid structure of written texts when compared to spoken discourse.
We should also note that common ground is not a static quantity. It is built up during the utterance. When we begin to listen to someone speaking, we form a hypothesis about what is to come on the basis of our knowledge of the person, previous discourse, and our own linguistic and world knowledge. Occasionally, we offer the speaker encouragement that we are following what they are saying (using a different track - Clark's term, 1996:255, 389-390 - from that in which we are conducting the main business), but in absolute terms we have to wait until well into a piece of discourse for our initial hypothesis to be confirmed. Indeed, we may change our hypothesis about how a given piece of discourse is going to end several times. Towards the end of the utterance, then, the possibilities for completing it will become fewer. We could look at this another way and say that the inductive probability (Cherry 1966:234) of an item at the end of an utterance is greater than that of an item at the beginning. Elements become increasingly redundant in the sense that a mistake at the end of the utterance will be more easily detected (Cherry 1966:186); by contrast, their potential information value is high, as substituting an unexpected element for an expected one will contribute relatively more information to the utterance than would be possible at the beginning. A confected example such as (1) makes this clear:
(1) It was a warm day in June, as the ice was just beginning to melt ...The second half of the sentence confounds our expectations that a day in June should be warm in our terms. Of course, the text will go on to place the utterance in the southern hemisphere, where June days are among the coldest. Our inference is clear: common ground and inductive probability are inversely proportional.
I now come to the nub of my paper: how are the phenomena of common ground and closure reflected in German word order? Of course, in 25 minutes you can't cover everything, so I shall try and be both selective and brief, confining myself to illustrations of the principle.
|VORFELD||FINITE VERB||MITTELFELD||NON-FINITE PARTS OF VERB||NACHFELD|
|Der Lehrer||regt sich||dauernd||auf,||weil die Studenten faul sind.|
Weinrich (1993:33) goes so far as to consider two-part verbs to be the standard. His sample data suggest that they outnumber one-part verbs in written text (1993:31-2), and, as he points out, all one-part verbs are potentially two-part verbs in any case, as they can all form a perfect, a future, past tense of KI, and so on. If this surmise turns out to be correct, the importance of the Nachfeld for the communicatively most dynamic elements, those not well grounded because they contain too much information, a matter which will be dealt with in greater detail below, is reinforced. This use of bracketing also provides an answer to the question of why the most strongly lexical elements should be placed towards the end of the VP, in the second half of the verbal bracket. In the case of what Weinrich calls the lexical bracket (1993:41-47), the non-finite element typically provides the contrast between a number of lexical items (e.g. sie gibt ... ab / acht / an / auf / preis / in Auftrag, etc.), and is therefore not well grounded. The same is true of grammatical bracketing (e.g. habe ... gespochen; bin ... gelaufen; werde ... antworten), in so far as the tense, the mood, even voice (a function of focusing), are all relatively well grounded in previous text, whereas the lexical element will probably be less likely to belong to common ground.
In terms of the development of German, I briefly note that the evolution of the Satzrahmen is a relatively recent phenomenon and is primarily a feature of the written language.
Bracketing is also a feature of noun phrases. The NP has conventionally been seen as having a Vorfeld and a Nachfeld, but no Mittelfeld, but it seems to me that this analysis misses useful parallels between NP and VP structure. If we analyze the nominal bracket as consisting of a determiner and a noun - the prototypical NP -, three positional fields also emerge. The Vorfeld could contain one of the predeterminers (all, irgend, manch, solch, welch, and so in colloquial language), but it could also be empty. The Mittelfeld could contain adjectives, typically inflected, and their dependent elements, always uninflected, and only in exceptional circumstances other elements such as a preposed genitive. The Nachfeld would then contain all other elements dependent on the noun, such as prepositional phrases, relative clauses and the like. The following table makes this clear.
|manch||ein||strenger||Blick||aus den Augen|
|--||die||beiden nächsten||Nachbarn||von uns|
If I may again mention as an historical aside that the development of nominale Blöcke is again relatively recent and a feature of the written language, but that the process as a whole probably postdates the development of the verbal bracket.
The last example in the NP table suggests a further bracketing phenomenon, that of prepositional phrases. Bracketing is most marked in the case of postpositions and circumpositions:
ersten Schätzungen zufolge
am Finanzamt vorbei
am Fiskus vorbei
an der Wirklichkeit vorbei
However, I have not had time to investigate this aspect of the problem.
Good evidence for the sequencing of firmly grounded elements over those which are less firmly grounded is supplied by the preferred order of elements in the Mittelfeld of German verb phrases. It has been common practice in descriptions of German word-order to distinguish several relatively homogeneous groups in terms of their relative positions in the Mittelfeld: unstressed pronouns; subject, dative, and accusative complements (some include the genitive in this group); the adjuncts; and the rest of the complements. It is also generally recognized that these groups are subject to internal ordering, so that a theory of word order based on common ground should be able to account for both the ordering of the groups themselves and any secondary sequencing which occurs within them.
(2) Darum habe ich es meiner Mutter gegeben. (Engel 1988:323) (3) In Taiwan überwacht die Crew die Entladung ihres Jets, dann führt sie der Dienst nach Hong Kong (Spiegel) (Durrell 1996:462) (4) Gestern hat es ihm Thomas geschenkt. Gestern hat er es seinem Freund geschenkt. (Duden 4:793)These are only capable of interpretation by reference to immediately preceding text or to the immediate domain of discourse: the personal pronouns refer to the participants, whereas the interpretation of third person pronouns relies on elements which have recently been established as common ground. Durrell's example (4) illustrates this principle nicely.
(5) Schließlich hat man meinen Eltern doch noch Briketts geliefert. (Engel 1988:323) (6) Gestern hat jemand meinem Vater eine Kettensäge geliehen. (Durrell 1996:463) (7) Vermutlich hat diese Briefbombe dem Politiker ein Rechtsextremer zugeschickt. (Duden 4:794) (x8) Gestern abend hat der Vater den Kindern eine abenteuerliche Geschichte erzählt. (after Weinrich 1993:80)In the light of common ground, this is natural, as definite NPs will typically identify entities which have already have been referred to in previous text, and which therefore belong to personal common ground, or identify entities which the speaker expects the hearer to understand without reference to previous text and which therefore belong to cultural common ground. Indefinite NPs, on the other hand, typically introduce new entities into the discourse.
(9) Wenn diese Staaten weit genug seien, sollen sie auch sofort in die Verhandlungen einbezogen werden. (dw, 25.10.1997)
(10) Die Unionisten forderten erneut, dass die katholische Untergrundorganisation IRA sofort mit der Entwaffnung beginnen müsse. (dw, 10 July 1999)
(11) Die Palästinenser beharren darauf, dass Israel sofort mit der weiteren Umsetzung des Abkommens von Wye beginnt. (dw, 7.8.1999)
In the light of common ground, this ordering principle is easily explained. Complements are determined by the verb and their incidence therefore belongs to the cultural common ground of German sentence patterns, whereas the incidence of adjuncts is determined by the speaker and can not belong to common ground until they have been expressed. A description based on functional sentence perspective would arrive at a similar view here, as a difference between complements and adjuncts in terms of their relative information value (compare Firbas' model of communciative dynamism, 1971) would be assumed. As complements are determined by the verb, their information value is relatively low. Adjuncts, on the other hand, often carry the most important information in an utterance, and their information value is relatively high.
The first piece of the jigsaw concerns obligatory orderings. Engel points out, for example, that estimating adjuncts never follow negating adjuncts (1988:334). In the light of common ground, we could argue that negating adjuncts always provide a contrast to what has gone before, and therefore contain information which the speaker knows the respondent knows is not common. Engel's examples illustrate the point (1988:334):
(12) Er hat es wahrscheinlich nicht gewusst. *Er hat es nicht wahrscheinlich gewusst.The second piece of the puzzle concerns the relative ordering of definite temporal adjuncts such as damals, früher, nächsten Freitag, and indefinite temporal adjuncts, such asimmer, bisweilen, noch and schon. The former tend to come first and the latter later in the Mittelfeld. Naturally, these elements seldom co-occur, so comparisons are only possible by substitution, in the following examples of (noch) immer by damals:
(13) Das war übringens nicht ihre Absicht.
(14) Davon wollte ich eigentlich nicht reden.
(15) Saddam habe bis heute den Verdacht nicht ausräumen können, dass seine Experten noch immer Massenvernichtungswaffen entwickelten.This conjecture must be confirmed on the basis of further empirical evidence. Certainly, this data does not prove that damals can not be placed after Experten, as it quite clearly can, merely that all other things being equal, this is the unmarked order, and that placing damals later in the VP implies that the speaker/writer recognizes that it is not well grounded.
--> Saddam habe bis heute den Verdacht nicht ausräumen können, dass damals seine Experten Massenvernichtungswaffen entwickelten.
(16) Bei dieser neuen und unbequemen Beschäftigung unterhielt sie ihre Einbildungskraft immer mit dem Bilde ihres süßen Freundes...
--> Bei dieser neuen und unbequemen Beschäftigung unterhielt sie damals ihre Einbildungskraft mit dem Bilde ihres süßen Freundes...
The most puzzling feature of the sequencing of adjuncts in the Mittelfeld is the tendency of estimating adjuncts, such as sozusagen, besonders, beispielsweise, bedauerlicherweise, to come towards the left in the Mittelfeld. On the face of it, as estimating adjuncts typically indicate the attitude of the speaker, the theory of common ground would predict that they should be placed later in the Mittelfeld. One explanation occurs to me, but at present I do not see how it could be verified. Accepting the theory as given for the purposes of argument, the effect of placing an element early in the Mittelfeld is to mark it as common ground. If a speaker wishes to steer the conversation, or a writer wishes to predispose the reader to accept their argument, the effect of marking attitudinal elements as common ground is to achieve acceptance on the part of the listener/reader. It could be compared to the rhetorical Anbiederungstaktik or Einschleichtaktik discussed by Engel (1988:95-7), which is used as a linguistic "preemptive strike" to overcome any resistance of the part of the text recipient. To this extent, it is possible to suggest that even the most puzzling feature of adjunct sequencing is not inconsistent with the theory of common ground. One could add that, even if this ordering is not an unconscious attempt to manipulate the recipient, but simply represents an open hand on the part of the speaker/writer, these attitudinal markers allow recipients to interpret the rest of the utterance in a way which allows for cooperation. The use of beispielsweise permits interpretation of the rest of the phrase as an example; erstens allows the listener to get ready for more points to come; bedauerlicherweise gives the recipient an opportunity to appreciate the speaker's regret.
I should like to talk about the Vorfeld and the Nachfeld as well, but time is against me, so I would like briefly to refer to the ordering of elements in the Mittelfeld of the NP.
It has been noted frequently, that adjectives are placed in a "natural", semantically based, order, usually described as a sequence of quantitative, referential, qualifying, geograohical and classifying adjectives (see Engel 1988:560, 635f.), as shown with examples in the table below:
|Quantitative Adjectives||Referential Adjectives||Qualifying Adjectives||Geographical Origin||Classifying Adjectives|
|cardinals and beid-, ganz-, viel-, wenig-, zahlreich- etc.||ordinals, damalig-, dortig-, erwähnt-, einstig-, obengenannt-, vorliegend-, seinerzeitig-||(most adjectives) alt-, blond-, frisch-, respektlos-, hölzern-,||Münchner, Berliner,||städtisch-, kirchlich-, staatlich-|
Keller's scheme (1978:584-6) further classifies the qualifiers into "general - colour - material - geographical origin - nominal", the last mentioned being those which are "especially closely linked with the noun", which appear to equate to Engel's classifying group. The examples Keller gives (1978:585) are
(17) die diesjährigen, erfolgreichen, französischen musikalischen Festspiele
(18) die zwei engen, grünen, wollenen, französischen Röcke
Semantic classes are difficult entities to handle, as the differing treatment of adjectives of geographical origin in Keller and Engel demonstrates. The categories quantitative and referential are elastic too. For example, are sämtlich-, ander-, weiter-, and others quantitative or referential? The adjective selb- is clearly referential but it functions almost as a demonstrative determiner (indeed, it becomes a determiner in some Swiss dialects, Keller 1961:62, e.g. säb Gschäft 1961:73). It may become increasingly problematic to propose such classes, as the following paragraphs show.
One may observe straight away that the sequencing as described in the handbooks reflects a cline from the general to the specific. In other words, adjectives less likely to limit the possibilities of following elements are placed earlier in the Mittelfeld of the NP than those with a potentially restricted collocational set. This may simply be a reflection of the general principle of ordering in German that, within a given field, elements tend to determine the elements to their right, in other words determinans comes before determinatum. Perhaps not coincidentally, this ordering is also the normal one in German compounds: Raubvogel is a type of bird, not a type of robbery (cf. Fleischer and Barz 1992:244ff.). German is, in Schmidt's terms (1961:10), an anticipating or regressive language. However, we should note that the sequencing represents an increase in the inductive probability of the elements, and therefore a decrease in their groundedness.
Referential adjectives refer to items in the domain of discourse, either in the text or in the context which is common to both participants. Indeed, they cannot be interpreted outside the context, so they must belong to common ground. In example (19), damalig- can only be interpreted with reference to Anfang 1995; in example (20), einstig- can only be interpreted with reference to 1994.
(19) In der vorangegangenen 13stündigen Sitzung war es vor allem um den Auftritt des Rechts-Extremisten Manfred Roeder an der Bundeswehr-Führungsakademie Anfang 1995 gegangen. Der damalige Chef des Akademiestabes, Oberst Norbert Schwarzer, sprach von einem "einmaligen Ausrutscher".Many similar examples could be cited. One problem with examples such as this is that quantifying and referential adjectives seldom co-occur. Another is that ordering appears to be driven by the following noun. In a corpus of over 850,000 running word forms, the adjective weiter- always came before the numeral when Personen, Touristen, Anlagen, Staaten, Städten and Mitarbeiter formed the nucleus of the NP (e.g. drei weitere Mitarbeiter, etc.); however, it came afterwards when Monate, and Arbeitsplätze formed the nucleus (e.g. weitere zwölf Monate, etc.). I have no explanation for this phenomenon at present. Further studies are needed.
(20) Die 1994 begonnene Privatisierung der Deutschen Lufthansa ist abgeschlossen. Von diesem Montag an werden sämtliche Papiere des einstigen Staatsunternehmens an der Börse gehandelt.
Neither is it true that referential adjectives always come before qualifiers. In the following example (21), the adjective tatsächlich- precedes damalig-, but why?
(21) Der letzte Innenminister der DDR, der CDU-Politiker Peter-Michael Diestel, muss sich vor dem Landgericht wegen Untreue verantworten. Diestel ist angeklagt, vor der deutschen Einigung im Juli 1990 von seinem Ministerium eine Villa weit unter Wert gekauft zu haben. Für das ehemalige DDR-Gästehaus nahe Berlin samt 3.500 Quadratmetern Seegrundstück soll er rund 190.000 Mark gezahlt haben. Den tatsächlichen damaligen Wert schätzen Gutachter dagegen auf rund 770.000 Mark.One answer might be that tatsächlich is more securely grounded than damalig. The former is dependent for its interpretation both on Untreue and weit unter Wert but damalig is dependent for its interpretation only on im Juli 1990. But there appears to be no way in which tatsächlich could ever follow damalig (and no examples have been found), so an alternative explanation may lie in the inherent meaning of the adjectives. On the one hand, damalig would seem to have a greater inductive probability than tatsächlich, in the sense that damalig normally refers only to events and states, whereas tatsächlich may refer to any entity. On the other hand, I note that temporal adjectives typically appear further to the right in the NP, as the following example (22) illustrates:
(22) MEKKA: Zehntausende von Panik ergriffene Moslems haben die bislang reibungslos verlaufene diesjährige Pilgerfahrt nach Mekka an ihrem letzten Tag in ein blutiges Chaos verwandelt.Clearly, further research on a larger corpus is needed here before secure conclusions can be drawn. However, the results obtained to date are not inconsistent with the notion of common ground.
This conclusion is supported by two general observations from the data. First, quantifying and referential adjectives almost invariably precede qualifying and classifying adjectives, as examples (23) to (27) show.
(23) außer diesen beiden notwendigen KleidungsstückenSecond, general qualifiers precede specific ones, as the following examples demonstrate.
(24) die beiden langen türkischen Flinten
(25) bei den vielen privaten Krankenkassen
(26) die dortige albanische Bevölkerungsmehrheit
(27) der damaligen sozialistischen Regierungspartei
(28) diese zwei schwarzen, mittels einer Kohle je auf der Mitte der Wange hervorgebrachten PunkteWhenever these basic patterns are violated, speakers and writers do this because they wish to achieve a particular effect, and this is usually clear from the context. It may be possible to make a distinction between ein seidenes enganliegendes Kleid and ein enganliegendes, seidenes Kleid as Keller (1978:585) suggests, but examples such as these are difficult to find in practice.
(29) Der neue Konzern wird mit einem Umsatz von etwa 65 Milliarden Mark und 190.000 Beschäftigten zu den zehn größten deutschen Unternehmen zählen.
As far as other Mittelfeld elements are concerned, the theory would predict that extended attributes, containing information assumed to be unknown to the recipient, would come later in the Mittelfeld than simpler elements, as in example (29) above.
Cherry, Colin, 1966. On Human Communication. Cambridge, Massachusetts and London: MIT Press. (2nd edn, first published 1957.)
Clark, Herbert H., 1996. Using Language. Cambridge: Cambridge University Press.
Durrell, Martin, 1996. Hammer's German Grammar and Usage. London: Arnold. (3rd revd edn.)
Engel, Ulrich, 1988. Deutsche Grammatik. Heidelberg: Groos.
Eroms, Hans-Werner, 1986. Funktionale Satzperspektive. Tübingen: Niemeyer. (Germanistische Arbeitshefte, 31.)
Firbas, J., 1971. On the concept of communicative dynamism in the theory of functional sentence perspective". In: Sbornik praci filosofike fakulty brnenske university A. 19. 135-144.
Hoey, Michael, 1991. Patterns of Lexis in Text. Oxford: Oxford University Press.
Keller, R. E. 1961. German Dialects. Manchester: Manchester University Press.
Keller, R. E., 1978. The German Language. London: Faber.
McCrone, John, 1999. Going Inside. A tour round a single moment of consciousness. London: faber.
Norman, D.A., 1988. The design of everyday things. New York: Doubleday.
Schmidt, K.-H., 1961. In: Indogermanische Forschungen 66.
Weinrich, Harald, 1993. Textgrammatik der deutschen Sprache. Mannheim, Leipzig, Vienna, Zurich: Dudenverlag.