Georg Bossong
MORPHEMIC MARIilNG OF TOPIC AND TOCUS
[(75) in : iviichei Kefer & iohan van cierAuwera (eds.)(i989). Universals of Language (:Belgian Journal of Linguistics 4),27-511
In theanalysis of basicsentence relations and constituents, afundamen- tal distinctionbetween cASE FUNCTToN and pnncmrrc FUNCTToN(henceforth CF and PF, respectively) seemsappropriate, at least as a heuristic instru- ment. Of course,such a binary, apparentlyclear-cut distinction can be misleadinginsofar as it is easily taken for a true picture of reality. The notionaldistinction between CF andPF is an instrumentfor analyzingthe basicsyntactic relations. If suchan instrument serves its purpose,it is useful and makesthe analysisclearer and more comprehensive;but at the same time,its instrumentalcharacter should neverbe forgotten. In everydaycom- munication, CF and PF are frequently bound together,thus forming mor- phosyntacticunits where the respective contribution of CF andPF is difficult to determine.But even then the differentiationis importantfor an exact understandingof the categoriesfound in individual languages.
PrototypicalcASE FUNcrroNs can best be described within a framework "dominates" of thepredicate-argument type: one central predicate or"binds" "dependent" a numberof argumentswhose number can vary betweenzero and perhapsfour or five. In most,perhaps in all languages,there exists a - morphologicallyand,/or syntactically definable - word classdistinction betweenvrr.n and NouNwhich is correlatedwith the predicate-argument domainin the following way: the primary functionof venss is predicativ- ity, the primary functionof Nouttsargumenthood. Goth NoUNSand vnnns can of course be modified in order to serve the oppositepurpose, if necessary.Broadly speaking,the domainof vnnnsin argumentfunction is subordination;the domainof NouNsin predicatefunction is the so-called nominalsentence.) cASE FUNCTIoNs form a classof semanticallydefined re- lationsbetween the core, i.e. thepredicate, and the depending a-rgument(s). Theserelations can be described on differentlevels of absrraction.Although discrepanciessubsist on terminology,the use of certainrather abstract lahls 28 GeorgBOSSONG for CF is fairly widespread.In this paper,the labels AGr, AG2, PAT1, and PAT" will be usedwithout further discussion(AG = AcENs,PAT = pArIENs;fhe subscriptnumbers refer to verbal valency).
IRAGMATICFUNcrIoNs are not directly connectedwith the semanticre- lations betweenpredicate and argument(s).They presupposethat thesere- lationsare in fact establishedat the sentencelevel, andthey serveto integrate the syntacticunits thus establishedin the discourse.They can be described numberof different, but in generalconverging parameters: old vs. new by a "setting" "core" information, startingpoint vs. finality of the utterance, vs. (Mathesius 1939:jddro), and so on. In this contribution, the differences between theseparameters will not be totally neglected,but they are not especiallyfocused on. What is focusedon insteadis theirconvergence, their common denominator.This commondenominator should be referred to by a pair of widespreadand theory-neutral terms. In this contribution, the terms roerc (TOP)and nocus (FOC) will beused for thispurpose. They cannotbe discussedhere in detail.Suffice it to saythat sentences in discourseusually containtopical and focal material,i.e. linguistic units which referto entities which are oldlknown from the context or from experience/previously mentionedin the contexVbackgrounded[inkedwith precedingunits in the text,etc. on theone hand, and linguistic units which areneVunknown to the hearer/notmentioned earlier/foregroundedlinked with following units in the text, etc. on the other hand.
In contrastto CFs, there is no dependencyrelation betweenPFs: neither does FOC depend on TOP nor vice-versa. Both are immediate constituentsof the sentence,in contrastto the casearguments which fill a framework previously establishedby the predicate.Of course,one and the samesyntactic unit in a real utterancecan be looked upon altematively from the CF and the PF perspective.
The expressionof PF differs greatlyfrom what is found with respectto CF. Evidently, the formal meansfor expressingpmgmatic relations are basically identical to thosewhich serveto expresscase relations, since the possibilities of formal representationof gtammatical meaning are rather limited in humanlanguage. In fact, the following setof binary oppositions necessarilyapplies to all instancesof formal expressionof grammatical categories:
Figure I implicit vs.;Elicit \ ta*e :'rc vs. xmmemlc ./\ partial vs. total TOPIC AND FOCUS 29
"Implicitness" means that the expression of grammatical categories is lacking, i.e. that the meaningof the categoryis implied. If two constituents, say M and N, are brought into a grammaticalrelation, this relation may be left implicit by the grammatical system of a given language. It can be deducedeither from the inherentmeaning of the constituentsthemselves, or from the specific discoursecontext ofa given utterance.A certain classof temporaladverbs in English (and in many other languages)may servehere as an example for illustration; in a sentencelike:
(1) Next week he'll comeÆIe'llcome next week the grammatical relation ADV ^ V is left unexpressed,in contrast to a sentencelike:
(2) He was born in 1948 where the preposition functions as a partial grammemicexpression of the samerelation (substandardFrench J'y vais à Paris would be an exampleof total grarnrnemicexpression). Note that the construction(1) is called "implicit" not only becauseof the lack of a preposition,but also becauseof the lack of a taxematicdifferentiation: the different placementof the adverb hasa bearingon the disribution of pragmaticfunctions within the sentence, but it doesnot alter the basicrelationship ADV ^ V.
Explicit expressioncan be purely taxemicor grammemic.This is best exemplified by the SUBJ ^ V and the OBJ Â V relation. (Note that the "subject" "object" traditional terms and are used here as abbreviationsof AG, & AG, and PAT, &.PAT, and thatthey arevalid only with respectto lanjuages df ttrenomihanve-accusauve type; non-accusatiïe languages are "&" not takeninto considerationin thiscontribution: the symbol standsfor categorialfusion.) A grammaticalrelationshipbetween two elementsM and N is said to be expressedtaxematically if M ^ N * N ^ M. This is the casein Chinesesentences like:
(3) wô ai nï vs. ni ai wô I love you you love me
(Note that the famous English exampleDog bites men vs. Man bites dog does not belong to the categoryof purely to(ematic expression:the verbal endingprovides a partial grammemicexpression which is combinedwith the taxemicone.) It is evidentthat the possibilitiesof taxemicexpression are rather limited; in a relation between two members, only two different categoriescan be expressed,as for instanceN ^ V for SUBJ Â V and V ^ N for OBJ Â V in the Chineseexample just quoted.For more differentiated systemsof grammaticalrelations, therefore,specific elementsare neces- 30 GeorgBOSSONG
"grammemes" sary; thesephonetically material elementsare termed here (Pottier l974,Heger 1976).This term is lessambiguous, less liable to be "morpheme", misunderstoodthan since this latter term can refer a) to grammaticalmorphemes (i.e. grammemes)and b) to any minimal meaning- "morpheme" ful unit in general.I prefer to usethe term in the b-senseonly. With specific grammemes,any number of categorial distinctions can be made since the number of grammemesis potentially as unlimited as the number of lexical items. However, prqtotypical grammemesare limited in number,and the fact that they form closedclasses is sometimestaken as a definitionalcriterion (lexemes, of course,form openclasses).
There are many possible criteria for classifying grammemic markers - criteria related to semantics,morphology and syntax. Among thesecriteria, the partial vs. totaldistinction is consideredhere as being the mostbasic and the mostuniversally applicable one. It follows immediately from thevery natureof grammaticalrelations: given a relationbetween two elementsM andN, thegrammeme can be addedeitherto one of theelements, or to both of them. Grammemic expressionof M À N can take the form "partial", pM ^ N M ^ qN, in which caseit will herebe called or the form / "total". pM ^ qN, in which case it will be termed Total grammemic "agreement" expressionis known as in somedomains of grcmmar. Note, however,that the notion of total grammemicexpression is more abstractand "agreement" thereforemore generalthan the term in traditional grammar andlinguistics. It appliesnot only to certainspecific levels of grammatical organization,such as noun and adjective, orverb andsubject, but to all kinds of grammaticalrelations in general.The Turkish genitiveconstruction offers a good illustrative exampleof the differentiation betweentotal and partial marking within a given relation (noun-noundetermination) in a single language:
(4) ev-in kapr-st vs. ev kapt-st house-GEN door-its house door-its 'the 'the door of the house' vs. housedoor'
It is evidentthat other criteriafor distinguishingsubclasses of gram- memesare important as well. For instance,it canbe relevantto distinguish betweenbound and free grammemes,between suffixation and prefixation, betweenagglutination and inflexion, and so on. In thepresent context, such distinctionsare of minor interest.
The main point to be made in this paper has to do with the degteeof grammaticizationof casefunctions in general,and the individual pragmatic functions in particular.At the outset,it is evidentthat PF'sare far lessgram- maticizedthan CF's.The most significantcorrelates of this high vs. low degreeof grammaticizationcan be summarizedin the following way: TOPIC AND FOCUS 3I
- CF is necessaryin all sentences.An utterancebecomes an utteranceby the presenceof CF. This statementis valid within individuallanguages asïell as in the perspectiveof cross-linguisticcomparison. - The reasonfor this seemsto be that the speechact of assertionis primarily bound to the expressionof CF. In contrastto CF, the expressionof PF is not obligatory in all sentenceswithin individual languages,nor is its formal expressionobligatory in a cross-linguisticperspective. - As will be shownin greaterdetail below, the expressionof PF dependson the degreeof pragmaticcontrast, i.e. the contrastbetween topical and focal elementsin the sentence.This degreeof contrastvaries geatly; thereis no neat yes-nodistinction, but rather a continuousscale with an unlimited number of intermediatevalues. From the point of view of CF, a given NP eitheris, or is not, a subject;from thepoint of view of PF it makessense to speakof high vs. low topicalityand/or focality. - With respect to CF, there is no clear predominanceof predicate over argumentor vice-versa;if we considerthe grammemicexpression of the OBJ ^ V relation, for instance,we find that partial expressionis frequent: the grammememay be addedonly to the noun (Chinese)or only to rheverb (Swahili); on the other hand,total grammemicexpression also occurs rather frequently (Basque).In contrastto that, in the domain of PF the marking of TOP clearly predominatesover the marking of FOC. - The degree of gtammaticization according to the criteria just outlined correlateswith the meansof formal expressionof grammaticalrelations as they are classified in figure 1. This figure can be consideredas the formal counterpartof the functional hierarchy of grammaticization:implicitness correspondsto the focal point of low grammaticization,total grammemic expressioncorresponds to the focal point of high grammaticization. - Accordingly, it canbe observedthat the expressionof CF is very rarely left implicit (e.g. Burmese);usually, it is explicit, either taxematicallyor grammemically. Total grammemic expressionis relatively frequent. (Of course,there are important differenceswith respectto the individual case functions,but this cannotbe discussedhere.) - On the other hand,the expressionof PF is usuallyleft implicit. If PF is expressedexplicitly, purely taxemic means are by far themost frequent ones. Grammemic expressionof PF is compilrativelyrare. - Given thepredominance of TOP overFOC,it is evidentthat in the caseof partial grammemic expressionit is the TOP and not the FOC that will be marked.Total grammemicexpression is extremelyrare (Quechua, Classical Arabic (seebelow)). - If it occurs, there is a clear difference betweenTOP and FOC insofar as grammemic marking of TOP is frequent or nearly obligatory, whereas 32 GeorgBOSSONG
grammemic marking of FOC is always optional and far less frequent in running text than TOP marking.
In this paper, the rare case of grammemic marking -of PF.will be considered.lwo groups of sample languageswill be analyzed in some detail: languageswittr partiat and with total grammemicexpression of PF. According-to *hat hai been stated in the preceding-Pg-aqaphs,partial gtam-eniic expressionin this casemeâns marking of TOP alone,-withthe éxclusionofpOC. T.od grammemicexpressionmeans markingof TOPand FOC. I do not know instancesof exclusiveFOC marking. Foreach of thesetwo groups,two geneticallyunrelated but geographi- cally and culrurally-chosen contiguous languagesplus one totally unrelated lan- guage have been as a sample-.For TOP marking, the sample éo*lprises Japaneseand Korean plus Pâez,a Chibcha languagespoken in Columbia. FôrTOP andFOC marking, the samplecomprises Quechua and Aymara plus ClassicalArabic. Specialemphasis is given to Japanese, Quechua,and Classical Arabic. Among linguists all over the world, the Japanese-p-ostposition-wa is probably th-ebelt-known example_qfgrammemic TOP marking-in any î-guugê. The oppositionof -wa (TOP marker) and -ga_(SURJmarker) has bee-nd'iscussed-intensely for the past two decades.It is impossibleto summarizehere the findingsand results of this extendedlinguistic discus- sion; I will only stress some points which are importalt fo1 1 qeel-er understandingof the relationshipbetween the marking of CF andof PF. The peculiarity of the -wa/-ga distinction in Modern Japanesecannot be fully understoodwithout a basicknowledge of its historicaldevelopment. There- fore, a few remarks on the function of these glammemes in Classical Japaneseare in order. First of all it must be stressedthat -8d in the function of SUBJ marker is a relatively recent innovation. In Classical Japanesethis grammeme servedas a marker of the genitive; -ga was not the only genitive marker: it was used side by side with the posçosition -no which is still in use in the modernlanguagê. The differencebetween -ga and -ng ls cgTPalable to the difference iound in the two genitive constructionsof Turkish (seeabove, example (4)): -no marks the determinativerelation in general,the deter- mining noun ranks low on the animacyhierarchy, and it is often usednon- refereitially; -g4 puts emphasison the determiningnoun whictr ranks high on the animaèy hierarchy and which is usually referential. The most characteristicdomain of -ga is the deicticpersonal pronoun' but it occurs also with personalproper namesand other animateand/or specific nouns. The distinôtionbetween the -ga andthe -za gpnitive seemsto be stmcturally similar to the distinction of markedand unmarked objects in languageswith Differential ObjectMarking (Bossong1985). Compare the following ex- ample(Lewin 1959:76): TOPICAND FOCUS 33
(5) Masamune-gakatana vs. Masamune-nokatana 'a 'Masamune'ssword' vs. Masamunesword'
Since the postposition -ga usually occurredin combination with animate nouns(which areof courseprototypical agents), and since it wasnecessarily usedfor marking the agentin subordinateconstructions (where the relation "lowered" SUBJ^ V is toDETÂN) itcametobe usedforagentsingeneral. In modem standardJapanese, -ga hasbecome a SUBJ marker.It should be notedthat there are dialects where -no hasbeen generalized instead, or where -no and-ga areused alternatively in the subjectfunction.
It is ofparticularinterest to our presentdiscussion that the construction N ^ no ^ V is not unknownin the classicallanguage in sentenceswhich we would translateby main clauses,i.e. in constructionswhich are not subor- dinate. Comparethe following example(Lewin 1959:202):
(6) aki=kaze-no fuku autumn=wind-GEN blow 'The autumnwind is blowing' "There A moreprecisetranslationwould be: is blowingof theaurumn wind". "Il In colloquial French,the sentencecould be renderedas y a le vent d'automnequi souffle".This constructionis a typical exampleof the so- calledthetic (as opposed to categorical)judgment(Sasse 1987): there is no pragmatic contrastwhatsoever, all differencesbetween TOP and FOC are levelled down. According to a widespreadopinion, in such casesthe sentenceas a whole hasrhematic value.
Usually,however, a subjectwhich is not specificallymarked as a topic doesnot takeany casemarker in ClassicalJapanese. This languageshows "normal" a markednesspanern which mustbeconsidered as in a typological perspective:the subjectis unmarked,all other casefunctions aremarked by specificgrammemes. The oppositionof SUBJvs. OBJ @AT, ADV, etc.)is morphologicallyprivative: 0 vs. -wo (-ni, -de, elc.).Two simpleexamples (intransitiveand transitive)will illustratethis point (Lewin 1959:201-203):
(7) tsuki akiraka-ni teru moon bright-ADV shine 'The moon is shiningbrightly'
(8) otôto mon-wo hiraku brother door-OBJ open 'Brother opensthe door' GeorgBOSSONG
All the constructionsexemplified in sentences(6) to (8) havedisappeared from the modernlanguage. The postposition-ga would be obligatory in all thesecases to mark the SUBJ function.
In contrastto CF, therewere no suchfarreaching changes with respect to PF. The FOC marker -wahad in ClassicalJapanese approximately the same functional range and syntactic behaviour as it has in the modem language.It must be stressed,however, that certainminor modifications of the marking systemhave taken place in the meantime.In the classical language,-14ldr was completely independent from the casemarking system: this posrpositionwas simply addedto the topicalizednoun phrase, regard- lessôf iti casefunction. Since the function SUBJ had the marker O, thisrule implied that a topical subjecthad -wa as its only ending.The OBJ was marked by the postposition-wo (today pronounced-o, but still written as
(9) ware sake-wa noma-zu I sake-TOP drink-NEG 'I do not drink sake' ("Sake I do not drink')
(10) seihai-woba ten-nl makase-mu success=or=failure-OBJ+TOPheaven-DAT entrust-INTENT 'As for victory or defeat,I will leavethe decisionto Heaven'
(11)oya-ni-wa musuko ni-zari-ki parent-DAT-TOP son resemble-NEG+CONIN-PAST 'The sondid not resemblehis parents'("To his parents,the sondid not resemble")
The main featuresof the systemin ClassicalJapanese may be summa- rized as follows:
-TOP (r2) \PF +TOP CF\ SUBJ -O (-no(-Sa)) -wa OBJ -wo -woba/-wa GEN -ga/-no DAT -ni -niwa DrR(...) -e -ewa TOPIC AND FOCUS 35
As for the thetic-categoricaldistinction, ClassicalJapanese exhibits a tripartite formal differentiation: there is a neutral form SUBJ-0 n Y, a marked thetic form SUBJ-no ^ V, and a marked categorical form SUBJ-wa ^ V. The postposition-no links the subject closely to the verb; -lrd separatesit from the verb very sharply.Consequently, sentences with -no subjectsare event-oriented,whereas sentences with -wa subjectsshow aclearpragmaticprofile: they areconstitutedby topic-orientedpredications. It is naturalthat nominal sentences (i.e. sentences with a nominalpredicate) always havea topicalizedsubject; compare the following examples(Lewin 1959:92,20I):
(13) dôsho-wa meishi-no ato-ni tsuku kotoba-nari verb-TOP noun-GEN rear-at follow word-is 'The verb is a word which follows the noun'
(14)nihon-wa shima=guni-nari Japan-TOPisland=country-is 'Japanis an archipelago'
In Modern Japanese,the tendencytowards setting off the basic functionsof SUBJand OBJ from all otherCF, a tendencywhich wasalready discemiblein the classicallanguage, was brought to its final consequence. The SUBJfunction is necessarilyexpressed by theposçosition -ga, and the OBJ function by the posçosition -wo (pronounced-o); in the caseof topi- calization, both theseposrpositions are replacedby the TOP marker -wa, whereasall otherpostpositions are combined with it. The new systemcan be summarizedas follows:
-TOP (ls) \ PF +TOP CF\ SUBJ -ga -wa OBJ -o -wa GEN -no DAT -ni -niwa DIR (...) -e -ewa
In contrastto ClassicalJapanese, in the modern languagethe explicit grammemicmarking of CF andPFhasbecome incompatible with respectto the basic syntacticrelations SUBJ and OBJ: the grammemesmust either distinguishAG andPAT, or markan undifferentiated TOP which canfullfil bothfunctions. The morphological oppositions have changed fromprivative to equipollent(-wc vs. -o and-wavs.-ga). This configurationdistinguishes Modern Japanesefrom the greatmajority of other languages.As for the use of theseforms, it wouldbe a simplificationto saythat -ga is theSUBJ marker GeorgBOSSONG in theticjudgments, and -wc theSUBJ markerincategoricalones. However, sucha siatementwould not be very far from the truth. It would probably be more adequateto say rhat-wa sentencesare always categorical, and that -ga Sentençescan representthetiC utterançes as well as neutral ones,where the thetic-categoricaldistinction is difficult to establish.The SUBJ slot is frequently empty; the subjectis usuallyleft unexpressedifrecoverable from the context. To a certain extent,the distinction of topical -wa subjectsand non-topical -ga subjectsis arbitrary;in many cases,the choice betweenthe grafirmemesis not governedby strict rules but left to the speaker'sdiscre- iion. Coyaud(1977) made an experiment whereby the place of theposçosi- tion in a contemporarytext on the history of chemistrywas left in blank and a number of informants were asked to fill the blanks oul The result is ambiguous:whereas in the majority of the cases,a cleartendency towards the useof either-wa or -ga canbe observed, unanimity is ratherrare. As one "on of the informantsput it: dirait que les gensjouent à pile ou face". Although sucha conclusionmay seemsomewhat exaggerated, it is never- thelessevident that the rules arefar from being compulsory.Topicalization is a shadeadded to the basic information carried by the verb-argument structure;even if its expressionis grammaticized,as it is in Japanese,the constraintsgoverning its useseem to be relativelyweak.
The correlationof -wawith categoricalnessand -ga with theticnesscan be illustrated by the following examplestaken from the text on the history of chemistryjust mentioned(Coyaud 1977: 85ff.):
(16)jikken=teki jijitsu-to kasetsu-to-no aida-ni experiment=alreality-and hypothesis-and-RLTinterspace-in ôku-no mujun-ga araware-ta much-RLT contradiction-SUBJappear-PAST 'Great contradictionsappeared between experimentalreality and hypothesis'
(17) subete-no kagôbutsu-wa nigen=tekikôsei-o motsu all-RLT compound-TOPtwofold composition-OBJhave 'All chemicalcompounds have a twofold composition' "to Cross-linguisticcomparison shows that verbs meaning appeat",or more generally"to comeinto being",ordinarily have a rhematicsubject, and that the sentenceswhere they occur form prototypical thetic judgments. A sentencelike (16) forms one singlepragmatic constituent. There can be no doubt about the non-topical characterof the subject; therefore,the use of -ga is compulsoryin the modernlanguage, and all 17informants of Coyaud agreewith the authorwith respectto the useof -ga.In contrastto that,(17) TOPIC AND FOCUS 37 has the typical bipartite constituencyof a categoricaljudgment; all infor- mants agreein this casein topicalizing the subjectby -wa.
It is impossible here to give a complete account of all the intricate problemsof TOP marking in modem Japanese.I wouldjust_liketo mention ône fact aboutword order.Japanese is known as being an SOV languageof the most rigid subtype.It is all the more remarkablethat in colloquial speech the topicalizedsubject frequently follows ùe rhematicpredicate. This is the afterthoughtpatternwhich is well attestedin numerousotherlanguages, too: becauseof its primary importancefor the speaker'smind, the focal elements come first in the sentence;the topic is thenadded for clarity's sake,but with a lowered intonationalcontour. This right dislocatedtopic must be followed by the topic marker -wa in colloquial Japanese.This construction thus provides an indirect proof (if necessary)that in languageswithout gram- memictopic marking,e.g. in colloquialFrench, such afterthought elements must be consideredas ordinary topics, and not as somethingelse. In the following examples,the French translationshows the similarity of the constructions(Kuwae 1980: 484):
(18)kirei desu-ne, ano ie-wa nice is-indeed that house-TOP 'Elle estjolie, cettemaison'
(19) yomimashi-taka, sono hon-wa? read-PAST Q that book-TOP 'L'avez-vouslu. ce livre?'
This structural convergencein two languageswith exactly opposedword order properties shows once more that the pragmatic rcgularities which influencethe orderof constituentsare independentfrom the typological laws and tendencieson the CF level: they areless subject to typological variation and more governedby universal tendencies(see Bossong 1980for a more detaileddiscussion).
Modern Korean behavesexactly like Modern Japanese:the topi- calizer replacesthe subjectand object postpositions, whereas it is combined with all otherpostpositions. As in Japanese,the opposition of SUBJand OBJ aswell asthe opposition of SUBJand TOP aremorphologically equipollent. The only difference betweenthese languagesis quite superficial: the Japanesepostpositions do not showany allomorphism, but theKorean ones do. ttris allomorphismis conditionedby the phoneticenvironment: the markersof SUBJ,OBJ, andTOP havedifferent allomorphsafter vowel and after consonant(symbolized here as x - y). In two out of three cases(OBJ and TOP) the relationshipbetween the allomorphsis synchronically quite obvious.The systemcan be summarizedas follows: 38 GeorgBOSSONG
(20) -TOP +TOP \PF CF' SUBJ -84 - -l -nun - -un OBJ -rûl - -irl -nitn - -ùn GEN -iti DAT -e(ge) -etge)nùn DIR (...) -ro - uro -ronun- -uronunv
A few exampleswill show the striking structuralsimilarities of Korean and Japanesein ùis domainfl-ewin andKim 1976:118):
(21) nae-ga sur-ùl masi-mnida I-SUBJ sake-OBJ drink-PolmE 'I am drinking sake' [neural description]
(22)na-nirn sur-iil masi-ji=anssii-mnida I-TOP sake-OBJ drink-NEG-POLITE '(As for me,) I do not drink sake'
(23) sur-ùn nae-ga masi-mnida sake-TOP I-SUBJ drink-POLITE 'Sake, I drink it' "subject" Kholodovié(1954:235) states that the takesthe form (n)in in the "subjects" greatmajority of cases; with -i - -ga arc significantlyrarer, and even more so are other postpositionsfollowed by the TOP marker. This seemsplausible, although no detailed statisticfigures are given. As in Japanese,the functionaldistinction of thetic and categoricaljudgments is fundamentalfor theformal distinctionof -i - -ga vs. -(n)ùn.Since categori- cal judgmentsare in generalfar more frequentin discoursethan thetic ones, the above-mentionedstatistic predominance of -(n)itnover -i - -ga canbe easilyexplained.
The third languagewith partial grammemicmarking of PF is Pâez,the mostimportant Chibcha language which is spokenby approximately44,000 peoplein the Caucaprovince in south-westernColumbia. The following structuralsketch is basedon Jung1984; in themeantime, Jung has presented a newversion of herdescription (Jung 1989) where several good illustrative examplesof the topic-marking constructionhave been suppressed. I prefer to quote the examplesfrom the older version of her work. The structureof TOP markingin Pâezis basicallysimilar as that found in thetwo East-Asian languagesdiscussed so far. The main differenceresults from the fact that Pâezis a languagewith DifferentialObject Marking @OM), wherebythe OBJ is markedby theDAT endingwhenever it is definite,but left without TOPIC AND FOCUS 39 any marker if it is indefinite.Definiteness is not equivalentto topicality, accordingiy,we find markedobjects with and without the TOP marker; on the other hand, it is obviously impossiblefor an indefinite object to be topicalized.There is an independentsubject form of pronounsand nouns which always endsin a vowel; this can be the original vowel of the stemor the vowel -o which is addedto the final consonant.(Indefinite objects lack thisfinal -a.)The TOP markerl' | (glottalstop) is addedto this SUBJending; theresult is a phoneticallyconditioned alternation of /'/ aftervowel, and/a' / after consonant.After anotherl'1, the endingtakes the form /sa'/; this latter variantis infrequentand will notbe mentionedin the following diagram.The TOP marker can be added to any sentenceconstituent. The rules can be summarizedas follows (sameconventions as above,plus  for variation accordingto DOM): (24) YF -TOP +TOP CF\ SUBJ Q--a -'--a' oBJ,8 @L: s - -a's -'sa'- a'sa' OBJpr @L -ry- ryi -rya'- -a'tya' GEN a DAT.s -'s- -a's -'sa' - a'sa' DATpr -ry-ryi -tya' - -a'tya' LOC(...) -te -te'
A few examples(from Jung 1984: 167ff., 185;cf. Jung 1989: 774) may illustratethese rules. Note that
(25)een-a yu'ptje-na u'j-ue-ts-na time-SUBJ change-ing go-IMPF-PROGR-3SG 'Times arechanging'
(26) wagas-a' tyl' nasa-yakj-a' puuty uy White-TOP DEM Pâez-with-TOPeach=other see we=fri-mee-ne fi'nze-' want=ed-NEG-3SGlive-HABIT 'White peopleand the Pâezlive without wantingto seeeach other'
(27)yat pand-na ùs-a' vs. yat-a's pand-na frs-a' house sweep-ing be-3SG house-OBJ' sweep-ing be-3SG 'S/he 'S/he is occupiedwith house-sweeping'vs. is sweepingthe house' 40 GeorgBOSSONG
(28) tyâ'-sa-'s-a' kim=yujva jii-me-a' DEM-ABS-OBJ -TOP nobody understand 'This isn't understoodby anybody'
(29) âch een-su-' tyâ'wë seena tjèy-sa ùs-a' today time-in-TOP so too hard-ABS be-3SG 'Today, there is too much hardship'
Sometimes,the topical subject is righçdislocated;as in colloquialJapanese, suchconstructions are found heredespite ofthe fact thatPiiez is basically an SOV languageof the rigid subtype(Jung 1984: 165):
(30) tyâa-ty ùus yajky-wa'j ji'p-da'u nasa-' DEM-OBJpI heart think-OBLIG must-lPl Pâez-TOP 'It is aboutthese [schools] that we musttake care, we the Pâez'
In mostvarieties of Quechua,there is onegrammeme -qa which marks the sentencetopic or the sentencetopics, and another grammeme -m(i) which servesto mark the focal elementin the sentence.Accordingly, in this languagethe expressionof both topic and focus is equally grammaticized. There is somedialect variation asto the form of thesegrammemes: -qamay berealized a s -ka or -xa; in mostdialects, -rni losesits final vowel if preceded by a vowel, keepingit if precededby a consonant,but in somevarieties the form -mi is preserved in all phonetic environments. In some southern dialects,-n is addedto -mi, or it replaces-rni. Among the varietieswhich I hâve examinedthere is only one which seemsto lack the focus marker -rni altogether(Olto, spokenin Amazonas/Peru).
The main differencesbetween the topic and the focus marker can be summarizedas follows: - There is one absoluterule: whereasthe topic marker can occur more than oncein a sentence,the focus markeris strictly limited to one occurrenceper sentence.This ruleis of coursea directconsequence of a universallaw: if we define the sentenceas the minimal independentutterance, it is evident that there must be exactly one assertivespeech act per sentence;on the other hand, the topics about which something is assertedcan be indefinitely multiplied,at leastin theory.It shouldbe noted,however, that, statistically, sentenceswith more than one topic are but a small minority. - Two other differencesbetween the two grammemesare not strict laws, but statisticaltendencies. First, it must be notedthat the topic marker is always relatively more frequentthan the focus marker. The frequencyof useof these grammemesmay vary geatly accordingto dialect,text class,and individual habits,but regardlessof thesevariations the topic marker always predomi- natesover the focusmarker. Secondly, although both the topic andthe focus TOPIC AND FOCUS 4l markers can be added to any sentenceconstituent, the topic marker is of course particularly frequent in combination with the subject; the focus marker displaysmore variability, but it is evident that it is more frequentin combination with objects, verbal predicates,and predicative nouns than with subjects.There is no one-to-onecorrespondence between pragmatic functions and casefunctions, thesedimensions are independentfrom each other; but there is a close affinity betweenTOP and SUBJ, and a lessclose affinity between FOC and OBJ and the other predicate-relatedsentence constituents. - Finally, the paradigmsto which thesegrammemes belong show different patterns.The TOP markeris the only memberof its class,whereas the FOC marker -rni is the basicelement in a paradigmof markerswhich compriseat least one more other member in most dialects; -mi is the basic element becauseit marksthe assertivespeech act in its prototypicalform: the speaker has witnessedthe assertedfact himself, or he is as sureabout it as if he had witnessed it personally. This FOC marker is directly opposed to -s(i/ (frequently palatalizedto -sài) which marks the assertionof facts known to the speakerby others:the speakerdeclares that he is not himselfresponsable ofthetruthvalueofwhathe is saying,butheunderlines thathe has only heard "reportative" aboutit. This grammaticalcategory, which will be termedhere (REP),isofcourse awell-known and frequently attestedphenomenon in the world's languages.What distinguishes Quechua from otherexamples of this categoryis the fact that in this languagethe REP markeris formally opposed to a basicassertive marker which explicitly expressesthe speaker'scommit- ment to the truth of what he is saying.Normally, if a reportativeexists in a given language,it is a part of the verbal conjugation,and it is positively markedoff againstthe basic assertion which is leftimplicit; in Quechua,both the basic assertionand the reportative assertionare marked explicitly by specific grammemes.It should be noted that the reportative marker is particularly frequentin traditionalnarrative texts when stories are told which could not have been witnessedby the narrator.Apart from basic -ni and reportative -s(À)i some grammariansdescribe certain other suffixes as "emotive" "focus markers", but it is doubtful that markerssuch as the -yd, "impressive" the -ma belong to the same category. These grammemes cannot be analyzedhere in detail.
Summarizingthe main pointsof the precedingdiscussion, one might saythatthe structureofQuechua shows clearly thatthetypological predomi- nance of TOP marking over FOC marking, which can be establishedby cross-linguistic comparison, is also valid within a given individual lan- guage.Furthermore, it shows a basic difference betweenthe syntagmatic and the paradigmaticaxis of language:syntagmatically, the FOC is unique whereasthe TOP canbe multiple within a given sentence;paradigmatically, there is only one kind of topicalizing, but severalmodalities of focalizing, GeorgBOSSONG which are related with the different degreesof conviction of the speaker. There is one truth per sentence,but it can be assertedabout severaltopics; on the other hand,there is only one way of establishinga topic, but different forms of telling the truth.
The systemof markersof TOP andFOC works independentlyfrom the casemarking system.Quechua nouns are marked for caseby a setof suffixes which is cross-referencedfor SUBJ and OBJ by suffixes on the verb. The nominative is formally unmarked,all other caseshave their own specific marker; there is no Differential Object Marking, all objects being equally marked by the accusativeending -ta (originally -kta, a form which is preservedin the Huanca dialect). From the perspectiveof typological èomparisonof casemarking systems,Quechua is a representativeof the "normal" mosi frequentand structure,which is characterizedbya privative oppositionbetween ùe basiccase (here the nominative)and all the lest. If sùôtra systemis combinedwith a setof PF markers,the result is obvious;it can be summarizedas follows (dialectof Ayacucho):
(31) .. PF -TOP +TOP +FOC CF\ SUBJ A -qa -m(i) OBJ -ta -taqa -tam GEN -pa DAT -paq -paqqa -paqmi DIR (...) -man -manqa -manmi
(Syllable-final[ql is pronouncedas a uvular fricative; consequently,an endinglike -paqqadoesnot containa lengthenedconsonant but a combina- tion of uvular fricative and stop.)
Although they arevery frequentin normalrunning text, the PF markers are not obligatory in sentenceswith averbal predicate:the sentencemay be topicless,and/or the verb with its agrcementmarkers is sufficient as a focus in-itself. There is one type of sentence,however, where the useof both the TOP andthe FOC markersis compulsory:the nominal sentence without the copulaverb. If the subjectof a nominalsentence is explicitly expressed,the copula verbka- is omitted.The subjectmust be accompaniedby the TOP marker -qa, andthe nominal predicateis followed by the FOC marker -mi. In other words: in sentenceswithout a verbal predicate(i.e. without a valencyrelated framework of casefunctions), explicit grammemicmarking of TOP andFOC is necessaryand sufficient to constitutethe sentenceas an assertiveutterance.
Sinceboth topic andfocusare marked by grammemes,word orderis not "free" only but it is not primarily usedas a meansfor expressingpragmatic TOPIC AND FOCUS 43 functions as such. In many languages,there is a strict separationbetween grammemic CF marking on the one hand, and taxemic PF marking on the other hand.In a languagelike Quechua,where PF marking is grammemic, too, word order is freely available for other things. I have the impression (which hasto beverified - or falsified - by deepenedempirical research)that it is usedin Quechuaforexpressing emotional overtones. In living dialogue, the focal elementof the sentenceis very frequently found at the beginning of the sentence,and the topic, if present,is relegatedto its end.On the other hand,in traditional narrativesthe orderTOP ^ FOC is far more frequentand probablyùe dominatingorderof constituents.It seemsthat the availability "emotive" of word order had led to an increasedfrequency of the order FOC ^ TOP: sincethe grammemicmarking permits to distinguishnot only SUBJ andOBJ but alsoTOP andFOC, thetaxemic distinction of TOP ^ FOC vs. FOC ^ TOP is freely availablefor expressinga shadeof meaningwhich is provisionally termed here [*srne1ive1.
V/hetherthis term is adequatemust still beshown by a detailedanalysis of Quechuatexts. In any case,the following generalconclusion can be drawn.Two constituentswhich arein a gtammaticalrelation with eachother must necessarilybe brought into somelinear order, becauseof the linear characterof humanlanguage. Consequently, taxemic expression is always possible. On the other hand, grammemic expressionof a grammatical relationis not a logical necessity;if it is madeuse of, taxemicexpression becomesavailable for other things. A languagesystem may, but need not, take advantageof this availability. In the caseof the fundamentalsentence relations,there seems to be a hierarchywith respectto the useof grammemic marking:if thereis anygrammemic marking, it will applyto casefunctions, with pragmaticfunctions being expressed taxematically; if both casefunc- tions andpragmaticfunctions are expressed grammemically, then taxematics becomesavailable for still more subtle,less rigid distinctions(such as [+emotive]) which are probably never expressedin a fully grammaticized way.
A few exampleswill illustrate the use of TOP and FOC marking in Quechua.In the first place,some basic regularities will be exemplifiedby Ayacucho Quechua.In the nominal sentence,both markersare obligatory:
(32) hatun-mi wasi-qa big-FOC house-TOP 'The houseis big'
Although the order FOC ^ TOP is exremely frequent,the inverseorder is of coursealso found, especiallyin the caseof TOP contrast;in the following example,the TOP marker is addedto a place adverb: GeorgBOSSONG
(33) kay-pi-qa sumaq-ta-m wif,a-n kawsay-kuna this-LOC-TOP good-ACC(=Any;-FOC grow-3SG crop-PL 'Here, the crops are glowing well'
The TOP marker may occur severaltimes in the sentence;in the following example,it is addedto the time adverband to the subject.Note, moreover, that the negationconsists of the preverbalsentence adverb manain combi- nation with the verbal suffix -chuwhich servesto mark the interrogativeand the negative,i.e. the non-assertivemoods; it is highly instructiveto observe that thè negationadverb mana is practically always followed by the FOC marker: in negative sentences,it is the negation itself which attracts the focus,whereas the verbal predicateis Unassertedand therefore unfocusable.
(34) paqarin-qa tayta-qa mana-mhamu-nqa-chu iomorrow-TOP father-TOPno-FOC come-FUT+3SG-NONASSERT 'Tomorrow, my father will not come'
Both the TOP and the FOC marker can stand alone in a sentence.The following exampleprovides a minimal pair of different focusing, with no topic specificallymarked (Soto Rufz 1976:I17ff.):
(35) pay-pa allqo-n-mi kachu-ru-ra he-GEN dog-his-FOC bite-'unexpectedly'-PAST 'He wasbitten by hisddg!' vs, pay-pa allqo-n kachu-ru-ra-m he-GEN dog-his bite-'unexpectedly'-PAST-FOC 'His dog hasbitten him!'
The reportativefocalizer functions in exactly the sameway as -ni itself. Its usecan be exemplifiedby ùe following minimal pair:
(36) hamu-nqa-s paqarin vs. hamu-nqa paqarin-si come-FUT+3SG-REP tomorrow come-FUT+3SG tomorrow-REP 'He'llcome 'He'll cdmetomorrow (they say)'vs. tomdrrow(they say)'
As for textfrequency, nothing definitive can be said. There is still much researchto bedone. However, the count of pragmaticmarkers in atraditional niurative text may give an approximateidea of the distribution of the TOP andthe FOC markers.In 1l pagesof runningtext I found 80 occurrencesof -q4 but only 15 occurrencesof -mi (Uhle,Kelm andTrimborn 1968:25-35)' The reportative marker -si occurs only twice,.in the first sentenceof the narrativeand in a sentencewhich is quotedbelow (38). It seemsthat, once the reportative assertionmood is established,it is no longer necessaryto TOPIC AND FOCUS 45 insist upon the fact that the reportedevents are known by hearsayand not by personal experience.The category of the reportative is grammaticized formally, but not functionally: it is expressedby a simplesuffix which forms a paradigmwith the ordinary FOC marker,but its useis not obligatory. The following examplesare instructive for the text function of the pmgmatic markers (Cuzco dialect). First, the use of the reportative marker at the beginningof a narrativecan be observed; the focus is on the mainprotagonist which is introducedfor the first time in this sentence:
(37) hoq kontor-si apuesta-ta rura-sqa hoq atoq-wan one condor-REPbet tSp.l-ACC make-PAST+3SGone fox-with 'A condor,so it is told, oncemade a bet with a fox'
The following sentencesare takenfrom a long narrativepassage. When the fire-rain is mentionedfor the first time, it is accompaniedby a focalizer, in this casethe reportative-si; lateron, it is part of a theticjudgment; finally, it is referred to as a topic in a categoricaljudgment:
(38) nina=para-s chaya-nqa[...] cheqa-paq-minina-para fire=rain-REP arrive-FUT+3SG truttr-for-FOC fire=rain chaya-sia-sqa[...] cheqa-paq-mi nins=para-Qa arrive-CONT- PA ST+3 SG truth-for-FOC fire=rain-TOP chaya-ska-sqa arrive-'still' -PAST+3SG 'Fire-rain will come, I'm told. [...] Really, fire-rain is coming! [...] Really,the fire-rainis still falling!'
The following sentencesexemplify the positional variability of the marked topic; in two successivesentences, the predicateis marked by the FOC marker, whereasthe TOP marker is addedto the time adverb which may precedeor follow the verb; in anotheroccurrence of the sameverb, the FOC marker is addedto the time adverb,the verbal predicatebeing left without any pragmatic marker:
(39) (a) kunan-qa now-TOP
(b) miqhu-ru-lla-sqa-y-ki-f,a-n eat-'unexpectedly' -' only' -' immediately'- 1SG SUBJ-2SGOBJ- 'already'-FOC kunan-qa now-TOP GeorgBOSSONG
(c) kunan-mi ichaqa miqhu-pu-lla-sqa-y-ki-fla now-FOC but eat-'against'-'only'-'immediately'-lSGSUBJ- 2SGOBJ-'already' '(a) Now I'll kill you immediately,(b) andl'lleatyou up immediately,now. (c) But it's now that I'll eat you up immediately'
Aymara is geneticallyunrelated with Quechua,although the contact between the two languageshas been very close since immemorial times. This case is comparable in some respects to the relationship between Japaneseand Korean. As aconsequence oflong andintensive contacts, there is not only a great amount of lexical borrowing, but also of mutual grammaticalinfluence. Aymara resemblesQuechua in much the sameway as Korean resemblesJapanese. This meansthat, despitethe absenceof geneticrelationship, the grcmmaticalcategories are more or lessidentical in boththese language pairs. Frequently it is possibleto translateQuechua into Aymara, and vice-versa,not only word by word, but morphemeby mor- pheme. Of course,there are also differences,but on the whole the fwo languagesare strikingly similarin theirgrammatical structure. ln thedomain of the basicrelations, the most importantdifference is the fact that Aymara hasDifferential Object Marking whereasQuechua does not. As we have seenbefore, in Quechuaany object takesthe marker -ta; in Aymara, only animateand/or definite objectstake the marker -rz, which is identical with the obligatory dative marker. The structureof Aymara resemblesthat of Spanishin thisrespect.
The marking of PF works asin Quechua:there is a TOP marker- 7a and a FOC marker-wa.In the dialectspoken in thePeruvian province of Puno, from which most examples are taken, there is a morphonological rule comparableto what is found in Quechua:after most consonants,the final -c is preserved,aftervowel and nasal consonants it is deleted.As in Quechua, the TOP marker is frequent and may occur several times in a sentence, whereasthe FOC marker is restrictedto one occurrenceper sentence;it is less frequently used than the TOP marker. In a short narrativeof approxi- mately10 pages of runningtext (Porterie-Gutierrez 198 I ), I havecounted 7 I occurrencesof the TOP marker -ya and 18 occurrencesof the FOC marker -wa. These proportions are roughly equivalent to what has been found in Quechua.The rules are also very similar.A few selectedexamples will serve hereto illusrate this point. Nominal sentencesnecessarily have the FOC marker and the TOP marker, usually in this order (comparc(32)):
(40) p"a1si-ki-w uka quta-n-1 moon-'only'-FOC that lake-in-TOP 'The moon wason that lake' TOPIC AND FOCUS A1
The focus frequently precedesthe topic, but it may also follow it (compare (33)):
(41) pobre tiwula-1 apena-w mistl-iritajna_ poor tSp.l Tiwula-TOP with difficulty tSp.l-FOC comeout-PAST 'Poor Tiwula came out with geat difficulty'
The topic marker may occur more than once in a sentence(compare (34)):
(42) uka-ru-1 uka pul"a=puÀaacakana-1 wali winlat-iritajna that-DAT-TOP that cactus thorn-TOP much cover-PAST 'Moreover,he coveredhim thickly with that thorny cactus'
The negationmarker necessarily attracts the FOC suffix; asin Quechua,the negatedverb is followed by the non-assertivesuffix -ri which is also a question marker; moreover, the negative sufTrx-k(a) is added (compare (34)):
(43)xani-w iras-k+(a)-ti no-FOC takeout-NEG- ISG-NONASSERT 'I didn't takeit out'
ClassicalArabic providesanother example of grammemicmarking of both TOP and FOC. In this language,a cleardistinction is madebetween "nominal" "verbal" sentencesand sentences.The latter ones may truly "verbal" a fully conjugatedverb, but they differ from the tntly contain "verbal" sentenceswith respectto theorder of constituents: sentencesbgin with the verb, which is followed optionally by SUBJ, OBJ and/or other "nominal" nominal arguments; sentencesconsist of a sentenceinitial TOP predicate,which may be purely nominal (X is Y), or be composedof and a "verbal" a verb andits complement(s).It seemsreasonable to arguethat the sentencetype is a grammaticizationof the theticjudgment, whereas the "nominal" sentencepattern provides a gnmmaticized frame for categorical judgments:in the verb-initial sentence,the whole utteranceis rhematic, is no pragmatic foregrounding; on the other hand, the syntagmatic there "nominal" conFastbetween topical and focal elements is constitutiveof the sentence which can be described as verb-second. Normally, we are accustomedto languageswhere the categoricaljudgment constitutesthe "unmarked" "basic" or case;in ClassicalArabic, thetic judgments lie at the baseof the unmarkedsentence pattern, and categoricaljudgments aremor- phologicallyand functionally marked. This patternof markednessdistribu- tion ofPF is perhapsgenerally valid in verb-initiallanguages. "verbal" "nominal" The text frequencyof the andthe patterndepends "verbal" largely on the discoursetype. In narrativediscourse, the pattern 48 GeorgBOSSONG prevails whereasall kinds of argumentativediscourse show an inclination "nominal" iowards the pattern. When reading narrative prose, like for instancethe A rabian Nights,the orderof constituentsis V(S)(O) over pages and pages,without any topicalizationor focalization of sentenceconstitu- ents.Itls a continuousflow of successiveactions and events.On the other hand, scientific, or religious, or philosophical prose shows a much more "peaks" "valleys". "dramatic" contrastof pragmatic and "nominal" Formally, a sentencemay be marked by the order of constituentsonly; as in all other languagesconsidered so far, morphemic of PF is by no means compulsory. The order TOP ^ PRED is marking "nominal", sufficient in itself for marking a sentenceas i.e. categorical. However, gtammemic marking of PF is very frequentin ClassicalArabic' Here again, it can be observedthat TOP marking is far more frequent in running text than FOC marking. Both the TOP marking and the FOC marking grammemesare sentence-bound,not word-bound.These markers are not integrated in some morphological paradigm, and they are not cliticized.The only formal constraintis a syntacticone: both the TOP and the FOC markersmustbe placed at the beginningof therespective pragmatic constituent. The TOP marker behaveslike a verb in one respect:it governsthe accusativecase, as if it werea transitiveverb. This particle,which is'inna in Arabic, has preservedits verbal characterto a higher degtee than in Hebrew (hinne). It is not a true verb, however, but rather a kind of interjection.Its original meaningis perhapsbest be renderedinto English by the particle lo!. T\e correspondingFOC marker la-, on the other hand, behâvesrather like a sentenceadverb. Its original meaningmust have been "truly, somethinglike verily". 'innc Theparticle mustbe followed by anoun phrase of somekind. This can be a noun not accompaniedby a preposition,or a prepositionalphrase. Prepositional phrasesdo not change their form. Other noun phrasesare normally put in the accusative,regardless of their casefunction. It is also possibleto leavethe SUBJ= TOP in its original nominativeform if innais "lightened" 'in 'lightened' usedin its form (a constructioncalled mulaffaf in nativegrammars); in this case,the use of theFOC marker/a- is obligatory. This construction is rather rare, however. Normally, the marked topic 'inna consistsof inna + N^^^ or + PP (after a preposition,the noun is automaticallyin the gJnïive). The FOC marker la- canbe followed by any part of speechor sentenceconstituent. The mostfrequent cases are predicate nounsand predicateadjectives, but prepositionalphrases and conjugated verbsare also possible.
As for text frequency,a count of the first 100verses of the third suraof 'innc the Koran hasgiven the following results: occurs38 times, andla- 6 TOPICAND FOCUS 49 times;in 35 cases,it is the SUBJ which is topicalized,in 3 casesit is a PP; as for the focalizedpredicate, be it with or without the explicit FOC marker /a-, it is nominal in 20 cases,and verbal in 15cases. Once more, we find that grammemicTOP marking is more frequentthan grammemic FOC marking. Surprisingly enough,the figures found in this text passageof the Koran are not too far away from thosefound in Quechuaand Aymara narratives.The ratio of TOP : FOC marking rangesfrom approximately6 : 1 in Koranic Arabic to 5.3 : 1 in Quechuaand 4 : 1 in Aymara. Of course,none of the analyzedsamples is representative,and much more researchis necessaryin this domain before anything can be said with certainty.
A few exampleswill show someof the regularitiesof PF marking in ClassicalArabic. The following casesare the most frequent and important ones (finer subdivisionsare ofcourse possibleand necessary,but they are omitted here for reasonsof space):
(44) TOP,. (NPAcc)^ FOC.. (nominal) 'inna l-dina 'inda llâh-i l-'islàm-u TOP DEFART-religionby God-GEN DEFART-devotion-NOM 'The religion by God is devotion' (Q 3,20)
(45) TOP,- (Mn..) ^ FOC-. (verbal) 'inna 'âdam-a llâh-a $tafà wa-nûlr-an TOP God-ACC choose+3scPERF Adam-ACC and-Noah-ACC 'God haschosen Adam and Noah' (Q 3,34)
(46) TOP* (NPAcc)^ FOC*- (nominal) 'inna llâh-a la-huwa l-'azlz-u TOP God-ACC FOC-he DEF ART-almighty-NOM l-hakim-u DEF ART-omniscient-NOM 'God is almighty,all-knowing' (Q 3,63)
(47) TOP,. (NPAcc)^ FOC*. (verbal) 'a-'inna-kumla-ta-3had-ùna 'anna ma'a llâh-i FOC-2IMP-testify-PL that with God-GEN Q-TOP-2PL'âlihat-an 'ulrà gods-ACC others 'Do you testify that there are other godswith God?' (Wright 1874-5: II, 86) 50 GeorgBOSSONG
(48) TOP.. (NPN.M)^ FOC* (nominal) 'in hâ{-âni la-Éâhir-âni TOP.,,*rrrNpo,this-DUAL NOM FOC-sorcerer-DUAL NOM 'Thesetwo are sorcerers'(Wright 1874-5:-II,88)
(49) TOP.- (PP) " FOC,* (nominal) 'inna fî {alika la-'âyat-an la-kum TOP in that FOC-token-ACC for-you 'Thereinthere is a tokenfor you' (Q 3,50)
(50) TOP.- (PP) " FOC_- (verbal) 'inna bi-hi tu-q'à l-'umùr-u TOP by-him 3IMP PASS-cicatrizeDEF ART-thing+PL-NOM wa-tu-r'abu and-3IMPPASS-repair 'By him all thingsare healed and restored' (Nôldeke 1897: 41) TOPIC AND FOCUS 5I
REFERENCES
BossoNc,Georg. 1980. Variabilité positionnelle et universaux pragmatiques. Bulletin de la Sociétéde Linguistique 75, 39-67 . - 1985.Empirische Universalienforschung. Differentielle Objektrnarkierung in den neuiranischenSprachen. Ttibingen: G unter Narr. Covruo,Maurice.1977. L'emploi deWA et GA (étuded'un textescientifique: histoire de la chimie).In A. TVlodarczyk(ed.).Recherches en syntâxe(Travaux du Groupede LinguistiqueJaponaise, Université de Paris VII, vol. IV), 79-97.Paris: L'Asiatheque. lbcrn, Klaus.1976. Monem, Wort, Satzund Text. Tûbingen: Max Niemeyer. Jwc, Ingrid. 1984.Grammatik des Pâez. Ein AbriB.Diss. Osnabrûck: Ms. [newversion printedas a dissertation,Osnabrûck 19891. KHor-ooovrë,A.A. 1954.OËerk grammatiki korejskogo jazyka. Moskva: Izd. Literatury na InostrannykhJazykakh. Kr;w.a.r,Kunio. 1980.Manuel de japonais 2. Paris:L'Asiatheque. LrwN,Bruno.l959.AbriBderjapanischenGrammatikaufderGrundlagederklassischen - Schrifsprache.Wiesbaden : Otto Harrassowitz. and TschongDae Ktu. 1976.Einftihrung in die koreanischeSprache. Heilbronn: GustavScherer. M,c,rlssrus,Vilém. 1939.O tak zvanémaktualnfm ëlenénim véty. Slovo a Slovesnost5, 17l-r74. Nôroexr,Theodor. 1897 U9631. Zur Grammatikdes klassischen Arabisch. Wien [Darm- stadt:Wissenschaftliche Buchgesellschaftl. Ponrrnm-Gunmnrz,Lrliane.l980.tæs relations actancielles en Aymara.Amerindia 5, 7- 29. Porrmn,Bernard. 1974. Linguistiquegénérale-Théorieetdescription. Paris: Klincksieck. Sessr,Hans-Jiirgen. 1987. The Thetic/CategoricalDistinction Revisiæd. Linguistics 25, 5l l-580. Soro Rr'12,Clodoaldo. 1976. Gramâtica Quechua Ayacucho-Chanca. Lima: Institutode EstudiosPeruanos. Urlr, Max, Ande Krw, and HermannTnnûoRN. 1968. Vom Kondorund vom Fuchs. Hirænmârchenaus den Bergen Perus. Ketschua und Deutsch. Berlin: Gebr. Mann. Wrucrr, William. 1874-5.A Grammarof the Arabic Language.Second ed. l,ondon: FredericNorgate, 2 vol.