i ;.1,ijr! jri:j...li'' ljl

LanguageTypology and LanguageUniversals Sprachtypologieund sprachlicheuniversalien La typologiedes langues et lesuniversaux linguistiques An InternationalHandbook / Ein internationalesHandbuch / Manuelinternational

Editedby I Herausgegebenvon / Edite par Martin Haspelmath' Ekkehard Konig Wulf Oesterreicher' WolfgangRaible

Volume2,1 2. Halbband/ Tome2

Walter de Gruvter ' Berlin ' New York 2001

i*.' 1492 XIV. Typologicalcharacterization of languagefamilies and linguisticareas

107.The European linguistic area: Standard Average European

L lntroduction guagesshare structural features which cannot 2. The major SAE features be due to retention from a common proto- 3. Somefurther likely SAE features languageand which give these languagesa 4. Degreesof membershipin SAE profile that makesthem stand out amongthe 5. How did SAE come into being? is thus no min- 6. Abbreviationsof languagenames surroundinglanguages. There 7. References imum number of languagesthat a linguistic area comprisesQtace Stolz 2001a).In prin- ciple, there could be a linguistic area con- 1. Introduction sisting of just two languages(though this would be rather uninteresting), and there This article summarizessome of the main are also very large(continent-sized) linguistic piecesof evidencefor a linguistic area (or areas (Dryer 1989a).Likewise, there is no li rlltt' i. rili Sprachbund)in Europe that comprises the minimum number of structural featuresthat

llll'itt Romance, Germanic and Balto-Slavic lan- the languagesmust sharein order to qualify guages,the Balkan languages.and more mar- as a Sprachbund.For instance, Jakobson "Eurasian ginally also the westernmostFinno-Ugrian (1931) establisheshis linguistic languageslthese will be called coreEuropean area" on the basis of just two phonological I' r, languagesin this article). This linguistic area features,but of course an area that shares I is sometimescalled Standard AverageEuro- more featuresis more interesting.As will be l.r pean (abbreviated SAE), following Whorf shown below, Standard Average European ; (1941)[956: 138].The existenceof this lin- languagesshare over a dozen highly charac- guistic area is a relatively new insight (cf. teristicfeatures, so we are dealingwith a very I Bechertet al. 1990,Bernini & Ramat 1996, interestingSprachbund. {i f ,, Haspelmath1998, van der Auwera 1998,Ko- A linguistic area is particularly striking 1 ,ii nig & Haspelmath1999). when it compriseslanguages from genealog- While the parallels (like t ;',r, close syntactic among ically unrelated languages the South the Balkan (* til;Ir languageshave struck linguists Asian linguistic area fut. 109), or the since the l9th century and the existenceof Mesoamericanlinguistic area (+ Art. ll0)), a Balkan Sprachbundhas been universally but this is not a necessaryfeature of a accepted,the European linguistic area has Sprachbund.The Balkan languagesare all long been overlooked.This may at first ap- Indo-European,but they are from different pear surprising,because the membersof the families within Indo-European (Romance, Sprachbundare among the best studied lan- Slavic. Greek. Albanian). and not all lan- guagesof the world. However, it is easy to guagesof thesefamilies belong to the Baikan understandwhy linguists have been slow to linguistic area, so nobody questionsthe va- appreciatethe significanceof the similarities lidity of the Balkan Sprachbund(- tut. 108). among the core European languages:Since In the case of SAE, three entire branches most comparative linguists know these lan- of Indo-European(Romance, Germanic and guagesparticularly well, they have tendedto Balto-Slavic) belong to the linguistic area. see non-European languagesas special and However, here too it is clear that we are unusual, and the similarities arnong the not dealing with a genealogicalgrouping, European languageshave not seemedsur- becausenobody ever proposed a branch of prising.Thus, it was only toward the end of Indo-European that consists of precisely the 20th century, as more and more had be- thesethree families. On the contrary, Indo- come known about the grammaticalproper- Europeaniststypically assumea particularly ties of the languagesof the rest of the world, closegenealogical relationship between Itaiic that linguists realizedhow peculiar the core and Celtic (and sometimeseven an Italo- European languagesare in some ways when Celtic protolanguage),but Romance(the sole seen in the world-wide context. From this descendantof ltalic) is inside SAE, while the perspective.Standard Average European may Celtic languagesdo not belong to SAE. And "exotic even appearas an language"(Dahl sinceso much is known about the grammat- 1990). ical properties that Proto-lndo-European A linguistic area can be recognizedwhen must have possessed,it is fairly easyto test a number of geographicallycontiguous lan- whether an SAE feature is an Indo-Euro- 107. The Europeanlinguistic area: Standard Average European 1493

peanismor not. As was shown in Haspelmath varieties,but Basqueseems to show very few (1998),most of the characteristicSAE fea- of them. Somewhatfurther to the east.Geor- tures (also called Europeanismshere) are not gian in the southernCaucasus (and perhaps Indo-Europeanismsbut later common inno_ the other Kartvelian languages) shares a vations. surprising number of featureswith the core Thus, what needsto be shown in order to .. European languages.These impressionistic demonstrate that a structural feature is a \ statementsshould eventually be quantified. Europeanismis but since it is not clear how much weisht (D that the greatmajority of core European shouldbe attachedto eachfeature. this is iot languagespossesses it; straightforward. (ii) that the geographically adjacent lan_ All ofthe featuresdiscussed below are svn- tactic. guages lack it (i. e. Celtic in the west, or concern the existenceof certiin Turkic, eastern Uralic, Abkhaz-Adygh- morphosyntacticcategories. I am not aware of any phonological ean and Nakh-Daghestanian in the eist, propertiescharacteristic of the and perhaps Afro-Asiatic in the south): core European languages(cf. Jakob- son l93l: "do (iii) that the eastern Indo-European lan- 182; six por ne udalos'najtini odnogo guageslack it (Armenian, Iranian. In_ obsdeevropejskogo... poloZitel'nogo dic); and fonologideskogopriznaka [so far not a sin[le Europe-widepositive (iv) that this feature is not found in the ma- phonological feature lias jority of the world's languages. been foundl"). Perhaps phonologists have not looked hard enough,but at leastone ma- Par ilarly the last point is not easy to de- jor recent study of word prosody in Euro- monstrate for many features because there pean languageshas not found any phonolog- are still far too few representativeworld-wide ical evidencefor StandardAverage-Europein studies of grammatical structures, so to the (van der Hulst et al. 1999,especially Maps extent that our knowledge -4) about the world's I (but cf. Pisani 1969).A few eeneralizi- Ianguagesis incomplete and biased. we can- tions are discussedby Ternes(199-3), but he not be sure about the European linguistic finds that in most respectsEuropean lan- area. In this article, I will cite whatevlr in_ guagesare unremarkablefrom a world-wide formation is available, and sometimes I will perspective.Perhaps the only featureswortlr have to resort to impressionistic observa- mentioning are the relatively large vowel in- tions. ventories(no 3-vowelor 4-vowelinventories) The designation "core European lan_ and the relativelycommou consonantclus- guage" for members of,SAE is diliberatelv ters (no restriction to CV syllables).In these vague, becausethe European linguistic area respects,European languages are not average. does not have sharp boundariei. It seems Dut they are by no meansextreme either. possibleto identify a nucleus consistins of continental West Germanic languages(i.g. 2. The major Standard Dutch, G_erman)and Gallo-Romance (e. g. Average French, Occitan, northern Italo-Romancej. European features For this set of languages,van der Auwera In this section I will discussa dozen sram- (1998a:-824)proposes the name Charlemagne matical featuresthat are characteristic6f the Spr )und. Of the other languages, th6se core European languagesand that together which are geographically furtier from this define the SAE Sprachbund. ln each case I center also seem to share significantly fewer will briefly dehne the feature and sive a few SAE features, i. e. Ibero-Romance. insular examplesfrom SAE languages.Thin a name Scandinavian(Icelandic and Faroese), East map, which indicatesthe approximate loca- Slavic(Russian, Ukrainian, Belorussian) and tion of languagesby the arrangementof (ab- Baltic. Even English, a West Germanic lan_ breviated)language names, shows the distri- guage, is clearly not within the nucleus. Of bution of the various feature values within the non-IndolEuropean languages of Europe, Europe. In eachcase it can be observedthat the-westernUralic languages(i. e. Hungarian the nuclear SAE languagesare within the and Balto-Finnic) are at least marginal-mem_ SAE isogloss, and that the marsinal lan- bers ofStandard AverageEuropea-n; they are guagestend to be outside the isoiloss to a ln many ways strikingly different from east_ greateror lesserextent. (part of thE material ern Uralic. Maltese also exhibits a number of presentedhere was already included in Has- Europeanisms not shared by other Arabic pelmath1998.) linguisticareas t494 XlV. Typologicalcharacterization of languagefamilies and

with relativepronouns 2.1. Definite and indefinite articles 2.2. Relativeclauses clausefound in languages Both a definite and an indefinitearticle (e' g' The type of relative Frenchor Russianseems to English the bookla book; 'tut' 62) exist in suchas German, -Romance Average European all and almost all Germanic lan- be unique to Standard by the follow- guagesplus some of the Balkan languages languagis. It is characterized clauseis post- Greek, perhapsAlbanian and Bul- ine four features:The relative iUoaern relative pro- garian), but not outside Standard Average no-minal, there is an inflecting the relative Enropean. To be sure, their forms and syn- noun, this pronoun introduces pronoun functionsas tactii behavior show considerablediversity clause,and ihe relative head'srole (see Nocentini 1996 for an overview)' but a resumptive,i. e. it signals the (cf. Lehmann 1984: their very existenceis characteristicenough' within tlie relativeclause English,a rela- The distribution of articlesin Europeanlan- 103-109, Comrie 1998)'In woman guagesis shown in Map 107.1.(Abbrevi- tive construction like the suspicious these fea- itions of languagenames are givenin the Ap' whom I describedalso displays all pendix.) tures. Furtherrnore,in most SAE languages interrog- f; rtttt' the relativepronoun is basedon an rl. ative pronoun (this is true of all Romance,all Mod- t- 'i': Slavii and some Germanic languages, Greek, as well as Hungarian and Geor- 1. li ern gian). (Languageslike German, whose rela- EU Dut p.onoun is basedon a demonstrative,or ll l Iiu. Finnish, which has a special relative pro- Fr Tal tr noun, are not common.) The geographical distribution of the relative pronoun strategy

Sftl tz\ is shownin MaP 107'2. CB lr Trk Am f :r' t :ii - definiteand indefinite article present - - - present ! it'ir, - onlydehnite article \ ririlil Map 107.1:Definite and indefinitearticle Eng Ddt Pol RG In large parts of eastern Europe there are Gm Cz HnE Uk no articlel at all (East Slavic, West Slavic, Fr Turki-c' Sln Finno-Ugrian other than Hungarian, It SCr Nakh-Daghestanian,Ikrtvelian). Some neigh- boring non-SAE languagesdo have definite articles(e. g. Celtic, Semitic,Abkhaz' Mord- vin). and Turkish has an indefinite article, - relativeclause with introducing relativepro- but no neighboring non-SAE languagehas articles.The only both definite and indefinite - - - - only particlerelative clause exception among Germanic languages,Ice- Map 107.2:Two relativeclause types in Europe landic (which only has definite articles like nearby Celtic), is also the most peripheral The only other type that is widespread in Germanic language geographically.We can Europe is the postnominal relative clause also be certain that the existenceof definite introduced by a relative particle (Lehmann and indefinite articles is not an Indo-Euro- 1984:85-87), which often occursin the same peanism: The Iranian and Indic languages languagebeside the resumptiverelative pro- i,au. g.n.tully lacked articles throughout no,in tlp" just described(an Englishelam.nle their history. -the radio that I bought)' Particle World-wide, articles are not nearly as would'be of this type exist in most Slavicand common as in Europe: According to Dryer's relatives "it languages,as well as in Scandina- (1989b:85) frndings, appearsthat abouta Romance unA Modern Greek, but also itti.d of the languagesof the world employ vian languag.s Irish (Lehmann 1984:88-90)' articles"(125 out of a sampleof about 400 in Welstian-d particle is sometimesdiffrcult.to languages).Only 3l languagesof those in The relative from a degenerate resumptlve Drler'J sample(i. e' lessthan 8%) have both distinguish in many European languages definite and indehnite articles. pronoun, and 107. The Eulolxan lirrguistic ateir: Statrdarcl Avct':tgc Littrtlpcatt r 495 it developedfrom a relativepronoun thlouglr ln contlastto the languagesiust metrtioned. the graclual loss of iuflectional distinctions. iu Slavic. l:inno-Ugliau and Arrneuianthc However. this also rneans tlrat the relativc perlcct is usually baseclon a participial clauseloscs its specificallyEuropean tlavor. constrr"rctiouwith ln activc participlcand a bectruseparticle relatives are also ilttested coprrla (c.g. Finnislt ole-n see-nu! [be-lsc 'l widely elsewherein the world (e. g. in l'cr'- lcceivc-nrcr] havc rcceived').llungarian sian, Moderu l-leblew. Nahuatl. Inclonesian. seemsto lack a pcrfcctcompletcly. Irt sotuc Yorubar,and Thai, cf. Lehnrann 1984:85- Nukh-Daghestanianlanguages (e. g. I-ez-gian 97). and Goclobeli;. tlre pcrfcct is folrnecl on the l'Iowever. 1he relativc prolloun stfategy basis of the past convclb plus thc colrula. clear'lyis typically liulopcau. It is not founcl Geolgiancorues closcst to tlrcSAlr prototypc in the easterrrInclo-Europcan languages, irncl in that its transitivepcrl'cct is bascdon a pas- as Comrie (1998:6l) notes."relative clauscs sivc palticiple.but this is conrbineclwith thc 'have'. fonned usilrg the relative pronoult stlatcgy copulurathcr than the transitivevcrb ate quite exceptioualoutside Europc. cxccpt so (hat t.lrcpcrlcct has a qulsi-pirssivcst luc- as a recent tcsult'of the influcnce of Euro- [ulc, with thc agcnt in tlrc dativecase ('The 'Ihe 'l peall languages... relative pronoun lcttcr is-writtcuto-mc'. r'atlrcrtban havc- strategythus seemsto be a lcmarkable areal wlitten thc letter'). In Welsh.tlre ;rell'ectis typological fcatulc of Er:rcaclclresscd by {'r"rlthcr' lescarch.Il''hervc'-vclbs turn out to be typi- cal of Eulope. tlrat woulcl fit with tlre ten- clcncyof lJulopcan languagesto have norni- nativeexpeliencers in expclientialverbs (sec thc rrextscction).

t?g 2.4. Nominativccxperiencers 'l'hcrc cei arc tw() witysol'cxplcssing lrk cxpcricuccr illguntct'lts of verbs of sensation. '11nys'-perfects cl'uotiotl. Map 107.3: in Eulope cognition and pcrception:The exlrericncer 'fypological t496 XIV. characterizatiouof languagefamilies and linguisticarcas

may be assiurilatedto agentsand coded as Europeanis fairly clear:lndic languagesare "dative it nourinativcsubicct (e. g. / /iAcit), or it ntrv rvcll-kno*'n tor thcir subiects"of be assiurilatedto a paticnt or goal, so that expericncervcrbs, so agaiu the l'caturcis the stimulusargument is codcdas thc nomi- not genetic(see also Masica1976, especially native subject (e.g. 1/ pleuseslrc,). In Bos- Map 6. for the areal distribution of dative song's(1998) typology, the first typc is called subjectsin Eurasia and northern Africa). generali;ing,and the secondtypc is called (SeeHaspelmath 2001 for rnore discussion irtverting.Bossong studies the expressionof of experiential predicatesin European lan- ten common expcricntialpredicates in 40 guages.) Europeanlauguages. llc cornputcsthc rcla- tion betweeninverting predicatesand gener- 2.5. Participialpassive alizingpredicates, arriving at ligurcsbctween StandardAverage Europcan languages typi- 0.0 fbr English (where all predicatesare cally have a canonicalpassive construction generalizing)and 5.0 for Lezgian(wherc all (* Art.67) formed rvith a passiveparticiple predicatesare invertingl. By arbitrarilydivid- plus an intransitivecopula-like verb ('be', 'becorne', I 1ll!l ing the languagesinto those showing pre- or tl.relike). In this passivethe :lill dorninantgeneralization (ratios betwecn0.0 original direct object becomesthc subjectand "': and 0.8) and those shorving prcdourinant the original subjcctmay be ornitted,but it inversion(ratios betrveen0.8 and 5.0). we may also be expressedas an adverbialagent arrive at the geographicalpattcrn shown iu phrase.Such constructionsoccur in all Ro- Map 107.4. nlauceand Germaniclanguages, but also in

l:irno87) S.rm(il.8l) N(r() l:) Swdo I L\lru.83)

I-tv(t s) Lil(ori:r) fl .ii IJB()0) t)ut (o fl) I'ol(0.881 t llrt(01{) Cnn(t;rt Cz(o 76) Fr(ou) I lngon) t;,ili &il(0.10) It(r{8) tliiii,, SCr {}ri sl'nor:t Ilonr(2.b) I f{iiiti, I'rt(olJl Blg()s) Ulilllr

Crktr:zlt Trkru

Map 107..1:Prcdontinant gcrrcrrlization (cel)ter) vs. inversion(pcriphcry)

Thus, Bossong'sstudy basically confirms all Slavic(including East Slavic) and Balkan earlier claims (Lazard 1990:246-47, Dahl languages,as rvell as in lrish. The geographi- 1990:7) that the gencraliziugtypc is charac- cal distributionof suchparticipial passives is teristic of SAE. although some of the fig- showniu Map 107.5. uresare perhaps a bit surprising(e. g. the tact that Hungarian turns out to be more SAE than Germanor Dutch, and the inclusionof Turkish, but not Romanian or Albanian. with respectto tirisfeature). It is not possible Ir Lt! to explaineverything here, but rveevidently ht Fog Dut Pol RE have before us a fairly typical pattem SAE Brt Cnn Cz with French and English at tl'recenter, Celtic ljr I lng Lkt (plus Icelandicthis tirne)at the rvesternmar- 5ln gin, Balto-Slavic,Fiuno-Ugrian and Cauca- It SCt siatrat the easternmargin, and lairly gradual Spn Srd llun V.E Bl8 transitionswithin the macro-areas.No sys- l'rt Alb Nllt Crk tcmaticworld-wide studies havc beenmade. but at least the behavior of easternIndo- Map 107.5:I'articipial passives in Europc t491 l0?. The Europeanlinguistic area: StandardAverage European

further here.) In Haspelmath No passivesexist in Nakh-Daghestanian and be considered 3l verb pairs in 2l-J1n- in }iungarian, and passivesof different for- (1993),I examined that languages differ mal types are found in Turkic, Georgian, and guages and found inchoative-causatlvepalrs Armenian (stem suffrx), in Basque, and in greatly in the way 'get'-passive: 'Terry languagesate anticau- Celtic (cf. the Welsh got are expressed:Some preferring anticausativesto his hitting by a snowball' for'Terry got hit sative-protninent, causalive-promi- by a snowball'). Finnish and Irish have pas- causatives,while others are sivesofa different syntactic type: In this con- nent. lt turns out that anticausative-promi- of SAE' In struction, only the subject is backgrounded, nence is a characteristicfeature Romanian, while the direct object remains in its place. my sample, German, French, Participial passives are very rare in lan- Russian, Modern Greek and Lithuanian suasesother than Standard Average Euro- show the highestpercentages of anticausative a pairs (between 100% and 74'h of all i"ui. ln Haspelmath (1990) I surveyed verb world-wide sample of eighty languagesand pairs that do not belong to the third,-non- Euro- found that a passive exists only in the mi- directed, type). The percentagein the nority of the languages(thirty-one)' Of these pean languagesof my sample are shown in thirty-one languages,only four have.a pas- Map 107.6. sive iormed from a participle plus an intran- two of them are Euro- Fin sitive auxiliary and 47"k pean languages (Latin and Danish). The Lit most common formal type of passive is the stem sufftx (found in twenty-ftve languages)' Syntactically, the possibility of an adverbial agent phrase is also by no means universal, languages(La- Jng but it is characteristic of SAE 44% zard 1990:246). Rom lzg It must be admitted that the SAE status of 96% 40'/" this feature is less evident than that of the frrst two features because the eastern lndo- European languages also tend to have pas- 65'/" sives of this type. In fact, in my 1990 study' the two non-European languageswith parti- - 70- 100%anticausatives - - - - ciple-auxiliary passiveswere Baluchi (an lra- 50-70%anticausatives nian language) and Maithili (an Indic lan- Map 107.6:Percentage of anticausativepairs guage).Thus', one might say that this feature genealogical feature' is an Indo-European By contrast, Asian languagesshow much the Celtic languages and However, at least lower percentagesof anticausatives,prefer- non-SAE branches of Indo- Armenian, two ring causativesinstead (e. g. Indonesian:0'%. do not have such passives, and European, Mongolian:1l%, Turkish: 34%,Hindi/Urdu is a non-Indo-European language Maltese 35u/o,Lezgian: 40'%). An intermediateposi- such a passive(calqued from Italian)' with tion is occupied by the Finno-Ugrian lan- Europe (Finnish 47'%' 2.6. AnticausativeProminence guages of eastern Udmurt 46oh, Hungarian Muk) as well as There are three ways in which languagescan Georgian (68%) and Armenian (65'lu).In a expressinchoative+ausative alternations such 'get 'break study involving more languagesfrom Asia, as losVlose', (intr.)/break (tr')', Africa and Europe but less language-partic- 'rise/raise'.One is by means of a causative ular detail. Masica (1976)found a clear dis- derivation (- Art.66), i.e' a derived verb tinctive pattern for Europe: few causatives, based on the inchoative member of the al- heavy reliance on anticausatives(see espe- ternation, e. g. Mongolian xail-uul''melt (tr)', 'melt his Maps 2 and 3). ln a recentworld- . from xajl- (intr.)'. The second is by cially of 18 verbs from 80 languages, means oi an anticausative derivation, i. e. a wide study al. (to appear) report that in in- derived verb based on the causativemember, Nichols et 'change pairs involving inanimate e. g. Russian izmenit''sja (intr')', from choative-causative (i. most typical subtype)' izimi{ tchange (tr.)'. (The third type, in participants e. the generallyfavored worldwide which neither member is derived from the ihe causativeis only in Europe. other, i. e. non'directedalternations, will not and is strongly disfavored r498 XIV. Typologicalcharacterization of languagefamilies and linguisticareas Anticausative-prominence _ is not an Indo_ hyena ate turopeanism: the hare's fish'). This type is not Older lndo_Europeanhad a found productive in Europe at all. Conve.srty,'Outiuefx- causativeformation, which losi ternal possessorsseem to be very rare outside its productivity in the European b;.;;;; Europe (the only caseI am awire of is E;.. but continued to be produciive in .urt.in cf. Ameka 1996),so this is a u".y Indo-European(cf. the low .oburr"^i ngrr. of:jy"-un_ ample of an SAE feature. ticausativesin Hindi/Urdu). 2.8. Negativepronouns and lack 2.7. Dative of externalpossessors verbal negation In Kdnig & Haspelmath(199g) and Hasoel_ The areal distribution of negationin Eurooe math (1999),we studied '& the clistributionof has been studied in detaif by Bernini external possessorsin thirty European lan_ Ramat (1996) (see also Ramat & Sernini guages(- Art. 73). We found three main lan_ 1990).Here I. rvill single out just on. urp..i guagetypes in Europe: (i) those with dative ol negatlon.the cooccurrenceofverbal nisa_ ex.ternalpossessors. e. g. Gernran ,1ll!l Die Mutter tion with negativeindefinite pronounr. i dlr- t'uscttt dent Kind die Huure .The mother is tinguish rwo main types:(i) V + Nt (verb I titi washingthe child'shair., (ii) thosewith loca_ negative indefinite), e. g. German tiv_eexternal possessors, 'nobody Niernand ,'i e.g. SwedishNrigon kontmt ,soireone comes', and (ii) NV + NI l;,i briit armen pd honont (negated brokJK verb + negative indefinitet. e.e. arm (lit. on him)', and (iii) those that laci Modern Greek Kandnasdhen irxete ,noboiv externalpossessors and must express (lit. porrar_ not) comes'.A third, *l^.a tvp. rnintl sors NP-internally, e.g. Englislr. me SAf be distinguished in which verbai n.nuiion teature.external possessors in the dative,is cooccu.rswith negativeindefinites onl/when lound ln Ronrance.Continental West Ger_ the indefinite follows the verb but manic, Balto-Slavic, noi when Hungarian anO Salkan it precedesit. e.g. Italian Nessunoyiene ,no_ languages (Greek, Albanian). North .Not Ger- body comes'.but Non ho vistonesszro I t; manic and Balto-Finnic languages nave toca- have.seen.nobody'.For our purposeswe can tive external possessors,i. e-.th--ey t:,ii ura ,or"_ crassrlythls type as a subtypeof (i), V + NI. what.peripheral SAE languages f i.ii;, *itt, ..rp".t The Standard Average European typ. to this feature.The geographical L [iiili. distribuiion V + NI (cf. Bernini & Ramat f SgO:f S+,Has_ ls slrownin Map 107.7. pelmath 1997:202).It is found in French(if we disregardthe particle .,le),Occitan and all \rennantc tanguages.as well as (in the mixed vanety) in Ibero- and ltalo_Romanceand Al_ banian (but not in Romanian or other Bal_ kan languages). D\]t pol Ru The geographicaldistribu_ Cm Cz tlon of the typesis shownon Man 107.g. F. Hng Ukr Bq stn It scr Fin spn srd Ronr fft Lfu Pd Alb BtB Lir Nllt Grk Pol Map 107.7:Dative extemal possessor H"g

In the far west (Welsh.Breron. English) and ln tne southeastlTurkish. Lezgian) ol-Europe Rom I zo thereare languageswhich Brs aG)-" do*notfruu. .^iJr_ _.\\--l nal possessorsat all. The eastern Irk M lndo_Euro_ Map pean languagesKurdish, persian 107.8:.Lu1ey"_g.rlacking verbal negation and Hindi/ wlth a negativeindefinite Urdu alsobelong to thistype. Outside Europe a fourth rypeenjoys considerable popularify: "relation-usurping" All the eastern European languages (Balto_ the type. where'hepos_ "usurps" Jlavrc. Frnno-Ugrian. Turkic, Nakh-Daghes- sessor the syntacticrelation of the theexceprion of Georgianiand possessum(e. g. Chichewa,a Bantulanguage, llniol).wirh 'The the Celtic languagesin the west ihow the has hyena are the hare the fish' i"i .firJ NV + NI type.This type is alsothat of the t499 107. The Europeanlinguistic area: Standard Average European are all at the west- eastern Indo-European languages (lranian The locativecomparatives (Breton) or the easterntiinge of and lndic), as well as that of the clear major- ern fringe Russian,Nenets. Ubykh' ity of the world's languages:Kahrel (1996) Europelf innish, other two types do tlot his studiednegation in a representativeworld- Turkish. Laz). The - the exceedcompara- wide sampleof 40 languagesand found only exist at all in Europe particularly in Africa. and the five languages with V + NI negative pat- tive is founcl occurs only in the terns, one of which is the SAE language conjoined comparative Dutch (the other four are Mangarayi (Aus- Americasand Oceania. Evenki, Chukchi (Siberia)'and Nama tralia), 2.10.Relative-based equative constructions (southern Africa)), as against 4l NV + Nl (equativeconstruc- patterns,and sevenothers. I found a very sim- Comparisonof equality than compartson ilar pattern in my (non-representative)sample tions) is discussedless often has undertaken-a of 40 languages(Haspelmath 1997: 202). of inequality,and nobody study of equativeson a world-wide scale' 2.9. Particlesin comparativeconstructions Still, there are good reasonsto think that Comparativeconstructions were investigated equativeconstructions provide evidence.for (Haspelmath& by Stassen(1985) in a world-wide study of l9 Sfandard AverageEuropean languages languages(- fut.75). Stassendistinguishes Buchholz1998). ln Europe.tnany that is based six main ways in which the standardof com- have an equativeconstruction parisonmay be expressed:Three kinds of loc- on an adverbialrelative-clause constructlon' 'bigger X'as Z itive comparatives('bigger from X', For example,Catalan has /or Z corn 'bigger and X is the to X', at X'), the exceedcomparative as X' (where Z is the adjective rela- ('Y is big exceedingX'), the conjoined com- standard).Catalan cont is an adverbial parative ('Y is big, X is little'), and the par- tive pronoun, and tan is a correlativedemon- is found ticle comparative('bigger than X')' The par- stratlve.A very similar construction (Portugrteselrio Z ticle in this latter type is often related to a elsewherein Romance relative pronoun (cf. English thanlthat' Latin contoX, Occitan tan Z cornal'), in Germantc quamlqui),and the casemarking of the stan- (German so Z tt'ie,f . in Slavic (Czechtak Z in Ro- dard is not influencedby the particle (so that jako X, Russiantak(oi) le Z kak .l,), 'I (olvan Z it is possibleto distinguish love you m-ore mani (katle Z sar l). in Hungarian (niin and in than she' from'I love you more than her'). mint X), in Finnish Z kuin X), fu Heine (1994)notes, the six typesare not Georgian(isetive Z rogorc-l'). ln the English of cs evenly distributed among the languages.of constiuction, the relative-clauseorigin but the world. Of the l8 particle comparatives is not fully transparent synchronically. in Stassen's\ample,l3 are in Europe, and of diachronicallyas derivesfrom a demonstra- the l7 Europeanlanguages in the sample,l3 tive (eull slld > a// so) that was also used lan- havea particle comparative.The distribution as a relativepronoun. In some Balkan guages. correlative demonstrativeis not within Europe again conforms to our expec- the 'as tations: Particle comparativesare found in used (e.g. Bulgarian xubaw koto lebe is Germanic, Romance, Balto-Slavic,the Bal- pretty as you'), but the standardmarker (There kans,Hungarian, Finnish and Basque,so this clearlyof relative-pronounorigin. is is the SABtype. The distribution is shown in probably some connectionbetween the rela- Map 107.9. iiue-p.onounorigin of equativemarkers and the relative-pronoun origin of comparatlve standardmarkers that we sawin $ 2.9.). Non-SAE languageshave quite dift'erent equativeconstructions. Many SOV languages in eastern Europe have a special equative standard marker (Lezgian -ti:, Kalmyk iitrg: also Basquebe:ain and Maltese daqs), and the Celtic languageshave a special (non- ,'uby demonstrative)marker on the adjective(e. g. " t", Z le X'EQUATIvEZ with X'). ln '. lrish chornh 'equ- Ttk the Scandinavianlanguages, the word (e. - particle comparative ally' is used on the adjective g. Swedish - - - - locative comparatlve lika Z som X'equally Z as X'). The distri- equative con- Map 107.9:Comparative types in Europe bution of the relative-based 107. The Europeanlinguistic area:Standard Average Europeurr r 50l , in theeastern Caucasus, and indee.d in manv striking, but which nevcrthelcsssecnr good otherparts of the world,but they*uy nru.", .have candidate-sfor Europeanisms.No maps-rvill hadsubject person agreement mirking.) be given for theset'eatures, ancl the .ui.l.n.. 2.12.Intensifier-reflexive differentiation will be summarizedonly briefly. Intensifiersare words like Englishself, Ger- 3.1. Verb fronting in polar interrogativcs mansclbsl, French nftnrc and Russiin ,ralr ln thc lllgc rnl.ioritv that characterize ol' luuguirue.s.folur irr- a noun phraserel.erent as tcllogativcsart centralas opposed urar.kedby interrogativein_ to an implicitor explicit tonation or illl interrogativc periplrery(e. g. The pope particli or both hir)selfgow i, i,i (- Art. 77). ht his audience,i. e. not just sanrpleof' 79 languages. the ior*ainolsf- Ultan (1978)tbund only ,.u.n tu,rtu,ft"i Art.57: Kcinig& Siemund1999). In manv showing the altcrnative srrategy oi uJ,.U Ianguages,the intensifie,.^pr.rrion i, ;i;6 fronting (oftcn called,'subject_i,erbin.",er_ used.as a reflexivepronoun, for inrtan." in Persran(xod-ai 'himself': !lon"). Of these.six are European(English. 'Hushang Huiang xocl_oi Rornanian. Russian. himself', and !pn9h, Hungo-ii,,,r, Huiang xoias_ru2did Finnish;the sevenrhlanguage [Hu-shangself-ecc saw] ,Hushing is lufafiyj. i" saw hini that the SAlr statusof verb self). However,a feature fronting ieenrs that is"typicai of beyond doubt. In lirct, the SAElanguages is the differentiation large rn,ri-o.ityoi oi..ff.^_ Germanic, Romance and Slir-victa,rguog., lvepronouns and intensifiers(Konig & Has_ (plus. peln'-'h Mo.dcrnGreek) appear to hauJ virb 1999).For instance,German'nas srclr trontlr'tg (refir rn polur qucstionsirr one for.nror ".ve)vs. selbst (intensifier), Russian tras another. sebja The three Europeaulanguages lor vs.sall, Italianhas si vs.rre$o, Greck which ltas Ultan explicitly reports thit rio verb eaft6vs. idtrjos. Map I 07.I 2 ,f,o*, if,.l"n_ rrontlng_occurs guagesin are pcr.iplrcral:Basque. Girc_ Europewith specialrcflexive pro- Irc and nouns Lithulnian. l:urtlrcrrrrorc.Sn E lu,.r_ that arenot identificalto intensifiers. guages arecharacterized by tlreabsence of an interrogative particle. In UItan.s Ootn, tir. nine .Europeanlanguagcs exhibiting u'p,ir_ trclc in polar qucstionsarc all pcriphcriilto a greateror lesscrextcnt: Ilasque, Irish, Scot_ tish Gaeiic, Albanian, Hungarian, Litlrua nian. Brt Russian,Finnish, Turk]sh (and I can add Nakh-Daghestanian).Vcrb fronting iir u6g polar questionswas suggesteclas a Eiiro_ peanism already by lleCkman ( 1934) (c.f. sPn Rm Lzg, Dahl 1990). Alb Blg c.8 Trk Am 3.2. Conrparativemarking of adjectives Map 107. l2: Intensifier-reflexivedifferentiation Most Europeanlanguagcs have special forrns lor ac|cctlvesoccurring in cornparativecol.l_ Intensil'rer-reflexive structlons. differentiationis not an For instancc. En-clishuscs thc ropeanism, sulllx -cr^in (The lldo becauseeastern Indo- .thisway dog-isbigg_er tltr-,,t Europeanlanguages have the ,ua. .*o..._ Iltecat). Such an inllcctionalmarkei of adiec_ sronlor intensifiersand reflexives(e. g. per_ trvesis not cotlrntouin the world'sIancuaces sian .r-or/-ai,Flindi aap).There are nJ oJ_ outs.ide^of_ Europc. Somc lan guagcs rri roirl. Irshed world-widestudies yet, but it seenrs l{lnd ot adverbialparticle nrodilying the ad_ that non-differentiationii very .ornrnon Jecrlve1'rnore'). but perhapsthe most cont_ aroundthe world, and while diffeientiation is mon type is representedby Japanese.rvhcre also foundelscwhere, it is not io"n,l i,i;;;;; the comparativescmantics is iar.iea Uy tlre unmedrately adjacentto Europeanlanguages. starrdardrnarker alone (e. g. itru_gu,,rki ,.u,li ookli .rhJ [dog-sunrcat from big] dog is Uig_ ger than the cat'). l. Somefurther likely SAE Special features ^ comparativeforms are found in all Germanic. n this Balto-slavic and Balkan lau_ section,I will mentiona few features (wirh vhich qy3Sel the exceptionof Ronraniauand areless well-documented than those iu Albanran),and rnost Rornlncc languages :2,or whosegeographical distribution is iess preserveat leastfour suppletivefonni (elc. r 500 XIV. Typologicalcharacterization of languagefamilies and linguisticareas

structionin Europeis shownin Map 107.10' in the world's languages,and strict subject following Haspelmath & Buchholz (1998: agreementis characteristicof a few European 297\. languages,some of which happen to be well- known. In her world-wide sample of 272 Nnts languages,Siewierska (1999) finds only two strict-agreementlanguages, Dutch (an SAE Kom language) and Vanimo (a Papuan language Guinea). Siewierska further notes Eng Dut Pol R6 Udm of New Gm Cz that outside of Europe, she is aware of only Fr Llng Uk Tat two additional strict-agreementlanguages that Sln are not in her sample (Anejom and Labu, two It SCr Oceaniclanguages). Gilligan (1987) reached LzB a sample crg a similar conclusionon the basisof of 100 languages.The distribution of strict subject agreementmarkers in some European i'''tll!t Map 107.10:Relative-based equative-constructions languagesis shownin MaP 107.11. tiil equatives :r:. Impressionalistically,relative-based ', to be rare in the world's languages,and Fin ,': seem Est the easternIndo-European languages do not Lfu ! seem to use them in general (however, a ur counterexampleis Punjabi). Pol @u* I 2.11. Subjectperson affixes as strict Hng LJk agreementmarkers eq 9n It scr The majority of the world's languageshave Sf Srd Rrn "lte: " bound personmarkers on the verb that cross- Prt Alb Bl8 GB Alm refer to the verb's subject (or agent). When Mtt Grk Trk thesesubject afftxes cooccur with overt sub- - [6nguagsswith strict subject agreement ject NPs (full NPs or independent subject - - - - languageswith obligatory subject pronouns' pronouns), they are called agreementmark- lacking verb agreement ers.However, in most languagesthey can oc- Map 107.11:Obligatory subjectpronouns cur on their own and need not cooccur with overt subjectNPs. For example,in the Bul- 'you The map showstwo non-contiguousareas in phrasevie rabotite (pl.) work', we garian which subject agreementsuffrxes cannot have subjectsufftx -ite (2nd person plural) seethe a referentialfunction: Germanic and Gallo- with the independent subject cooccurring Romance languageswith Welsh on the one pronoun lie 'you (pl.)', showing that -ile is hand, and Russianon the other.Perhaps only an agreementmarker. But in Bulgarian it is the western European area should be thouefit equally possibleand probably more common of as being relevant for SAE; in Russian, to sayjust rabotite'you (pl.) work', i.e. the past-tenseverbs do not have subjectperson subjectsullix can have a referentialfunction affrxes,so Russianis not a very good exam- on its orvn. In German, by contrast. this is 'you ple of a strict-agreement language. In the not possible: work' \s ihr arbeit-el.Since easternNordic languages(Norwegian, Swed- the agreementsufftx -el does not have such ish, Danish), the subjectpronouns are obliga- an independentreferential function, the sub- English, German or Ice- ject pronoun rlr cannot be omitted' Lan- tory as they are in "non- havelost agreement guageslike German are often called landic, but the languages distinctionson the verb entirely (cf' Swedish pro-drop languages", and languages like 'llyoulhe "pro-drop jag biter bite(s)', Bulgarian are called languages"; biterldu biterlhan "strict-agreement biturlhann bitur). T\ese better terms would be lan- Icelandic 69 bftbrt "non-pro-drop" in a guages"vs. "referential-agreementlanguages". languages are thus lan- It has sometimesbeeu thought that strict sense,but they are not strict-agreement as agreement,as exhibitedby German, English, guages.English is approachingthis type' and French, is the norm and that referential the only remnant of subjectagreement is the -s. agreementis somehow special. But in fact, 3rd person singular present-tensesuflix referentialagreement is far more widespread (There are also some languagesof this type

r!;l,i :,na I 502 XIV. Typologicalcharacterization of languagefamilies and linguisticareas

there are Italian maggiore'bigger', ntinore'smaller', However, within the and-languages sub-typesaccording to the position of peggiore'worse',migliore'better'). Compara- several "A (e.g. haundi- the particle, which we may call and-B", tive forms also exist in Basque "A-and "A "A-ind 8", B-and", and B-and" ago'bigg-er'),Hungarian (nagy-obb'bigg-er')' "and-A Finnish(iso-mpi'bigg-er'), and other Finno- (of the remaininglogical possibilities' "and-A and- Ugrian languages. B" seemsto be inexistent,and pattern).Most Comparativeforms are not completelyun- B" occursonly as a secondary languages, and in particular all known outside of Europe. Arabic has a spe- European "A belong to the sub-type cial comparativefbrm (e.g. ?akbar'bigger', SAE languages, "A- 'big'), "A-and and from kabiir but it is unique among and-B". The types B-and" of the Afro-Asiatic languagesin this respect.Old and B" are found in some languages languagegas Indo-Iranian languages had comparative Caucasusand in some Turkic northern Eu- forms, and the modern Iranian languages well as scattered throughout (e. fuchi, havepreserved them to someextent (e. g. Per- rasia and South fuia g. in Abkhaz, sian-ter. Zaza -tr\. But further east,in mod- Persian.Sinhalese, Tamil, Burmese,Korean 'lll to Stassen;Stassen also points out I l ern Indic, the comparative does not exist according anymore,and languageslike Hindi-Urdu and that there is a correlation with verb-final llil periph- Bengali use a constructionanalogous to the word order here). Furthertnore' some use Japaneseexample just cited. Similarly,in the eral Europeanlanguages make restricted (e. my s toboj Uralic languages,the further east we go, the of the ruitft-strategy g. Russian 'I 'we also Old fewer comparativeswe find. For instance' and you', lit. with you', and Khanty (a Finno-Ugrian languagespoken in Irish, Lithuanian, Polish and Hungarian, these westernSiberia, i. e. outside of Europe) does accordingto Stassen).Taken together, "A and- havea comparativeform in 'sak (e.g. iant-sak data do show that belongingto the 'better'), which is used when no standard is B" type is not a trivial feature of the SAE present.But in a completecomparative con- linguistic area. : struction. no marking is found on the adjec- tl 3.4. Comitative-instrumentalsyncretism tive (e.g. narl ke:se:-ne:x'elt iarn [you knife- i, 'better preposition that I 2sc from goodl than your knife'. Ni- In all SAE languages,the (: also f:',rt, kolaeva 1999:21). expressesaccompaniment comitative) l.;:,1 (e.g. Thus, although this featureis not contined servesto expressthe instrument role ham- h;;rl, to Europe, it is typical of a SAE feature in English with: w'ithher husbqndlrith the languagesare said to exhibit com- I rl:ttli; that it is robustly present in western [ndo- rrer). Such Stolz (1996) EEIIIi European and Uralic languages,but gets itative-instrumentalsyncretism. rarer the further eastwe go in thesefamilies. studiedcomitative and instrumentalmarkers in a world-wide sampleof 323languages and "A 3.3. and-B"conjunction found that this kind of syncretismis typical more The feature discussedin this section is less of Europe. Non-European languages possessseparate markers for these distinctivethan the others mentioned so far' commonly 'with but I hope to show that it is not at all devoid two semantic roles (e.g. Swahili na (instrumental)'. As of interest. Stassen(2000) offers the first (comitative)', krvc'with of Stolz's world-wide typologicalstudy of NP conjunc- Table 107.1shows, about two thirds and only tion strategies,based on a sample of 260 samplelanguages are non-syncretic, (The l1n- languages(- Art.82). He distinguishestwo one quarter is syncretic. remaining which I ig- basic types, and-languages(using a symmet- guugesbelong to a mixed type, the ric particle) and l'illr-languages (using an nore here for the sakeof simplicity;thus, asymmetriccomitative marker). Two thirds percentagesdo not add uP to 100%.) from the of Stassen'ssample languagesare and-lan' Two areas diverge significantly guages,and sinceSAE clearly belongsto this generaltrend: Oceaniahas far lesssyncretism has far type, too, it is not a very distinctiveproperty. ihan the world average,and Europe average' And-languagescover all of northern Eurasia, more syncretism than the world Europe' South Asia, the Middle East and northern When we look at the pattern within we are dealing Africa, Australia, New Guinea, and parts of it becomeseven clearer that (as cf' Central and South America. Mllr-languages with an SAE feature Stolz recognizes, are encounteredin sub-SaharanAfrica, East 1996:120). Of the l6 non-syncreticlanguages i.e' and SoutheastAsia, the islands of Oceania, in Europe, l0 are Caucasianlanguages, one is and large areasof North and South America. they are clearly outside of SAE, and 107. The Europeanlinguistic area: StandardAverage European I 503

Table 107.1:comitative-instrumental: syncretic and non-syncreticlanguages

syncretic(e. g. English) non-syncretlc(e. g. Swahili) languages percentage languages percentage Europe 25 49% l6 3r,yn Africa , 20 3t% 38 58% Americas l6 2t% 54 69% tuia l2 18% 4t 7lo Oceania 6 t0% 54 86% World 79 24% 209 65"1,

only politically, not anthropologically, in This is clearly a very marginal feature in Europe (Greenlandic). Four of the remainine grammar, but it is intriguing that it should five languages are also otherwisenot typica-i show such a clear geographicaldistribution. instancesof SAE (Basque, Finnish, tr,tattese. Mari). And when we look at the 38 Indo- 3.6. Some other characteristicsof SAE European languagesin Stolz's sample, we The features examined so far present the seethat syncretismcannot be regarded as an most striking evideucefor StandardAverase Indo-Europeanism:Of the eight Indo-Euro- Furopean,but there are probably -uny -oi" pean languagesnot spoken in Europe, only featuresthat will turn out to be characieristic th show syncretism,while five show non- of the core European languagesin one way syncretism.Thus, in Asia Indo-Europearr lan- or another. In this subsection,several such guages behavelike Asian languages,and there candidateswill be mentionedbrieflv. The first is no generalpattern for Indo-European. few featuresin the following list ire purely negative:At first glance,this may seemodd, 3.5. Suppletive second ordinal but of course the lack of a category that is Most languageshave a suppletive form of the widespread elsewhereis no 'first', lesJ sienificant ordinal numeral i. e. a form not de- than the presenceof a category that is rare .one'. rived from the cardinal numeral An elsewhere. exampleis Gerrnan,where'lst' is erster(un- (i) Lack of an alienable/inalienableopposi- relatedto eins'l'), contrastingwith other tion in adnominal possession(' Art. jil. ln ordinals such as zweiter'2nd' (cf. zwei ,2'\, Nichols's(1992) world-wide sample. almost vierter'4th' (cf. vier'4'), and so on. In Stolz's half of the languagesshow such an opposi- (2001b)study of 100 ldnguagesworld-wide, tion, but no Europeanlanguage does there are 95 languages with special ordinal lillZ: 123).More generally,this opposition is rarer numerals, and of these, 78 have a suppletive 'hrst'. in the Old World and common in the New word for Thus, languages that say 'oneth' World, but in Europe it is evenless common (literally) for 'lst' are not common. than in Africa and Asia. However, the same sample has only 22 lan- .2nd,, (ii) Lack of an inclusive/exclusiveopposition guages in which the word for too, is in.first personnon-singular pronouns. Again, suppletive and not derived from..2' (e.g. this opposition is commonest in the New E1 ;h second).Thus, most languageshave 'twoth' '2nd'. World and in the Pacihc region, but in (literally) for The 22languages '2nd'word Europe it is even rarer than in Africa and that have a suppletive are lieavilv Asia, as was shownby Nichols(1992: 123). concentratedin Europe: 17 are European (iii) Lack of reduplicatingconstructions. I languages,and this type is clearly the mijor- have no systematicevidence to back uo the ity within Europe (which is represented by 27 claim that this is a characteristicfeature of languagesin Stolz's sample).Of the l0 Euro- European languages,but reduplicationis so pean languagesthat do not have a suppletive common acrosslanguages that its almost to_ second ordinal, six are clearly outside SAE tal absencein the core European languages (Basque,Turkish, Armenian, Georgian, Lez- becomesstriking. (Interestingly,reduplication gian, Greenlandic). Among SAE linguages, existedin older Indo-Europeanlanguages at only some Balkan languages(Romanian, Al- least in one construction, the perfect, but banian, Romani) and German lack a supple- even here it was lost entirely by the Middle tive secondordinal. Ages.) I 504 Typologicalcharacterization of languagefamilies and linguisticareas

(iv) Discourse pragmatic notions such as detail I have to refer the reader to van der topic and focus are expressedprimarily by Auwera'sthorough study. "Preterite sentencestress and word order difl-erences (ix) decay": the loss of the old (Lazard1998: I I 6). Only the Celticlanguages preterite and its replacementby the former and French give a very prominent role to present perfect. This is a change that oc- clefting, and particles rnarking discourse curredin the last milleniumin French,Ger- pragmaticnotions are virtuallyunknown. man and northern Italian, as well as in some (v) SVO basicword order at the levelof the other adjacentEuropean languages (cf. Thie- clause.This feature is of course found else- roff 2000: 285). Its distribution is far nar- where in the world, but in Europe it corre- rower than that of the other Europeanisms, lates particularly well with rhe orher SAE but it is the only feature of those studiedby features. The Celtic languagesin the west Thieroff whose geography comes close to have VSO order (exceptfor Breton, which is Standard Average European (cf. also Abra- also otherwisemore SAE than Irish and ham 1999). Welsh),and the easternlanguages have SOV word order. Interestingly,Balto-Finnic (Fin- Quite a l-ew additional features have been nish, Estonian,etc.) and (lessunequivocally) mentionedin the earlier literature as charac- Hungarian have SVO word order, whereas teristic of SAE, but earlier authors have sometimes theeastern Uralic languageshave SOV. Simi- neglectedto make surethat a pro- posed larly, the eastern Indo-European languages Europeanismis not also common else- where tend to show SOV word order. (See Drver in the world. Most of Whorf's original examples 1998 ior more on word ordcr in the lin- of SAE featuresseem to be of this kind. For instance,he notes that in contrast guagesof Europe.) "imaginary (vi) to SAE, Hopi lacks plurals" (such European languagestend to have just 'ten "meta- one converb(- Art. 83) (cf. Nedjalkov 1998). as days', according to Whorf a phorical For instance, Romance languageshave the aggregate").But of course,we have no gerundioIgt ronrlrf,English has the -lng-form, evidencethat such plurals of time-span nouns and Slavicand Balkan languageshave their are in any way characteristicof Euro- pean languages. adverbial participle. The Celtic languagesin It may well be that they are common the westcompletely lack sucha fonn, and the throughout the world. (To give Whorf his languageseast of SAE tend to have more due, it must be added that he was not interested than one converb. Otherwisethe core Euro- in demonstrating that SAE Ianguages just pean languagestend to have adverbial con- form a Sprachbund.He used this term as a convenient abbreviation for junctions (' Art.63) to make adverbial "English clauses.According to Kortmann (1997: 344\, and other European languages "a likely they have large, semanticallyhighly dif- to be known to the reader", without necessarily ferentiatedinventory of free adverbial sub- implying that theselanguages are ordinators placed in clause-initialposition". an exclusiveclub.) More generally,they tend to havefinite rather than non-finite subordinate (r strategies 4. Degrees of membership in SAE Art. 100), though a multi-purpose infinitive usually exists (except lbl the Balkan lan- Membership in a Sprachburrdis typically a guages). matter of degree.Usually there is a core of (vii) Europeanlanguages usually have a spe- languagesthat clearly belong to the Sprach- cial construction for negative coordination, bund, and a periphery of surrounding lan- e.g. English neitlrcrA nor B, Italian ni A nt guagesthat share features of the linguistic .8, Russianni A ni,B, Dutch nochA nochB, area to a greateror lesserextent. Hungarian sem A sem B. Again, no world- In order to quantify the degreesof mem- wide study has been published,but such a bership in SAE, a simple proceduresuggests negativecoordinating construction is rarely itself that was first applied to areal typology reported from languagesoutside Europe (cf'. by van der Auwera (1998a).In addition to Haspelmathto appear). individual maps in which the linesdenote iso- (viii) SAE languageshave a largenumber of glosses(as in Maps 107.1-12),we cancom- characteristicproperties in the area of phasal bine ditlerent featuresin a single map and adverbials(expressions like alread1,,still, no show the number of isoglossesshared by the "cluster longer, not yet) (van der Auwera 1998b). language.Map 107.13shows such a Theseare rather well documented.but for the map" in which the linesstand for "quantified 107.The Europeanlinguistic area: Standard Average European r 505 "isopleths"). isoglosses"(or The map com- historical role played by speakersof these binesnine featuresof$ 2.: definiteand indcfi- trvolanguages both in thc earlyrr.redieval his- nite articles,relative clauses with relativepro- tory of contiuentalE,urope and !n tlie vety nouns,'have'-perfect,participial passive,da- recentattempt at Eulopcan unillcation,this tive external possessors,negative pronouns is of courscau extrcrlelyintriguing result. and lack of verbal negation, relative-based (b) The southcrnEuropcan languages (both equativeconstructions, subject person afllxes Romanccand Ilalkan languages)arc at lcast as strict agreementmarkers, and intcnsitier- as closc,if not closcrto lhc nuclcusthan thc reflcxivediflbrentiation. The languagesin the northcrn languagesaud l-uglish.This means nucleus(French and German) show the SAI: that it is nrisleadingto call SAE lcaturcs "Westcrn value in all nine of thesefeatures. The lan- Iiuropcan lcatur.es",as is sonrc- guagesin the next layer (Dutch, other Ro- tirnes donc. It is truc that the Slavic lan- mance, Albanian) show eight features, tire guagesin the cast lack many SAE features. nextlayer (English, Greek, Romanian) shows but thc Balkanlangr.rirgcs ar.c gcncr.allv urorc sevenfeatures, and so on. In this map, the SAE than Slavic,altliough thcy are not wcst- resulting picture is actually very clcar, be- em European. causethe SAE area with at least five SAE, (c) Englanclstands sonrcwhat apart ll-onithc featuresstands out from thc remaininglan- Europcannuclcus (as notcd also by van dcr guages,which have at most two SAE fca- Auwcra 1998a:823), althou_qhit is closcly tures. related gencalogicallyto Gcrnran ancl has beenthoroughly influenccd by Frcnch.Sincc English is cr.rrrcntly I:in Nnls thc donrinantlanguagc --rL.t throughout thc rvorld, it is rvorth pointing out its sonte$,hatmarginal status among its Lt Euroltcansistcr languagcs. -1 Rrl Rustl Llttrn I It is inrpclltantto kcclt irt rnind that thc l'ca- uLr / T.rr turcson whichMap 107.13is basedhave nr.-rt becn sclcctcdrandomly and are thus by no mcansrcprcscntative. of the ntorphosyntactic . .. v.l: li'itturcsol' Luropcan languagcs.thcl' rvcrc crs I Irk Ann includedprecisely bccause thcy werc knowrr to showa distributionthat suppor.tsthe SAI: Map 107.13:A cluster map conrbiningnine fcl- hypotlresis.Thus. no clainr is rnadc turcs that all (or cvcnthc nrajorityol-) fcutures rvill shorvu siurilar distribution. It is perl'cctlypossiblc Suchcluster maps are thus a fairly direct rep- that rve u,iil some day discover anothcr resentationof degreesof membershipin a lin- Spracltbuntl,bascd on a difi'crent set of t'ca- guistic area. But of course, the cluster rnao tures,thal hasRussian at its coreand extends dircctly reflectsthe choiceof featuresthat arc all the way to wcsternSiberia in the eastancl combined,and this choice is always sorne- centralAsia in the south,but within EuroDe what arbitrary. Of the twelve featuresin g 2, conrpriscsouly the Slavic,Balkan. ancl onlv ninewere selected here bccause inforrna- Scai- dinavianlanguagcs. This areawould tic n the other three was incomplete.Icle- overlao with SAE. but it rvould not ally, the featuresof g 3 should have bcen contradicrii. Thus, a languagenray in principlebelong added,too. But it seemsto me that thc mairr to dill'crcnt linguistic arcas.aud dif]'cr-entlin- resultsof Map 107.13would uot be changed "on guisticarcas uray cocxist top (this rnap can also be compared to thc vcry of' eaclr othcr. Sinceareal typology is only in similarmap in van der Auwera(1998a: 823), its in- fancy, rve do not knorv how comnron rvhich combines five adverbial featurcs or such situaLionsare, but nothing l-eatureclusters). The ntost striking featurcs in the logic of a Sproclbuttd of Map 107.13are: implicsthat thc rvorldshould bc exhaustivelydivisible into non-overlapping (a) The nucleusof Standard AverageEuro- Sprachbtindc. pean is formed by French and German (a In fact, a nuurber of srnallerlinguistic hnding that led van der Auwera (1998a:824) areaswithin Eulope have bcen proposedin to proposethe term Clnrlemagne Sprachbund the literature(apart li'om the Balkan arca. lor the nucleararea of SAE). In viervof the rvhoscimportance is ltot doubtcdby anyonc). l 506 XIV. Typologicalcharacterization of languagefamilies and linguisticareas

e.g. by Lewy (1942).Wagner (1959). Decsy The first possibility must be rejectedbe- (1973),Haarmann (1976), and Ureland(1985) causethe great majority of Europeanisms (cf. also Wintschalek1993 on a Volga-Kama are innovationswith respectto Proto-lndo- area).Currently the most thoroughlystudied European.For instance,as far as we know, areasare the Circum-Balticarea (cf. Stolz Proto-lndo-Europeandid not havearticles, a 'have'-perfect."A 1991, Dahl & Koptjevskaja-Tamm(eds.) and-B" conjunction,strict 2001) and the Mediterraneanarea (cf. Cris- subjectagreement. particle comparatives,or tofaro & Putzu (eds.) 2000). However,no relative clauseswith relative pronouns (cf. strongclaims about a Circum-Balticor a Me- Lehmann 1974,Haspelmath 1998). With re- diterraneanlinguistic area seemto have been spectto Proto-lndo-European,and alsowith madeas a resultof thesestudies. respect to the oldest Indo-European lan- guagesattested in Europe (Ancient Greek, Old Latin, Gothic),Standard Average Euro- into 5. How did SAE come being'l peanis clearlyan innovation. The second possrbility,a pre-Indo-Euro- 'iil!l Linguisticareas arise through languagecon- tact, but preciselywhich contact situation pean substratumin Europe causingthe SAE riI gave rise to Standard Average E,uropeanis features,would be extremelydiffrcult to de- not inmediatelyclear. And what is the source monstrate,but it might be worth pursuing.It of the variousEuropeanisms: Who borrowed is intriguing to note that the geographical from whom? A full discussionof the socio- spaceoccupied by SAE languagescoincides preciselywith the areaof the Old Euro- historical,cultural and sociolinguisticissues lairly pean layer is beyondthe scopeof this article,so I will hydronymy,i. e. the homogeneous Hans Krahe restrictmyself here to mentioningjust hve of river names discoveredby (see possibilities: Vennemann1994 for recentdiscussion). Vennemann(1994) proposesthat theseOld (i) retentionof Proto-Indo-Europeanstruc- Europeanhydronyms were not coinedby an tures and assimilationol some non- early prehistoricIndo-European population, Indo-Europeanlanguages to lndo-Euro- but by a pre-lndo-Europeanpeople which he peanlanguage structure; calls Vasconic(the only surviving Vasconic (ii) intluence from a common substratum language being Basque). Furthermore, the of a pre-Indo-Europeanpopulation in Old European hydronymy is hardly attested Europe: in the British lsles, where the Celtic lan- (iii) contacts during the great trans- guagesare spoken,i. e. they could not have formations at the transition from late beeninfluenced by the Vasconicsubstratum. antiquity to the early Middle Ages in This is in perfectharmony with the well-mo- Europe; tivated hypothesisthat the Celtic languages (iv) the oftlcial language(Latin) and the acquiredsome of their striking t'eaturesfrom common Europeanculture of the Mid- a different substratum related to the Afro- dle Ages; Asiaticlanguages (Pokorny 1927-30, Gens- (v) the common Europeanculture of mod- ler 1993). ern times.liom the Renaissanceto the The main argumentagainst the substratum Enlightenment. view is that the SAE featuresseem to be gain- L ing ground too late for a pre-Indo-European The tlfth possibilitymust be rejectedbecause substratumto have causedthem. Some SAE a time depth of 300-500 yearsis not suffr- features appear only in the first millenium cient to accountfor grammaticalcommon- CE, but also the earlierfeatures usually come alities of the kind discussedabove. If lexical fairly late, so that the earliestrecords of Indo- similaritiesbetween the European languages European-languagesin Europe still show - are discussed tor instance neoclassical traces of the Proto-Indo-Europeanpatterns compounding (socio-lpaleo-lortho-ldemo-, (e.g. causatives,relative clauses, locative com- -graphyl-log1,l-cracy,etc.) or idiomatic struc- parative."A B-and" conjunction).If these ture (e.g. ivorl' to*'erltorued'avoriolElferrbein- SAE featureswere causedby a substratum, tunn, as poor as a clurc'lt tnouselpattvreconune then we should have much more evidenceof un rat d'igliselarnr v'ie eine Kirchennraus)- the population speakingthis substratumlan- then the last severalcenturies are the appro- guage.Moreover, a Vasconicsubstratum can priate time frame lbr explainingthe historical hardly account for the SAE featuresbecause links. but the basicsyntactic structures com- modern Basqueis in most relevantways very mon to SAE languagesmust be older. much unlikethe SAE languages. 107. The Europeanlinguistic area: Standard Average European I 507

Of the remaining two possibilities,we can 6. Abbreviationsof languagenames probablyexclude option (iv) (tne influenceof Latin in the Middle Ages),because most SAE Alb Albanian featureswere absentin Latin and developed Arm Armenian only in the Romance languages.There are Blg Bulgarian only two featuresfor which Latin influence Brt Breton is a likely factor: negation and relative pro- Bsq Basque nouns.In the caseof thesetwo features.the Cz Czech standard languagessometimes show devia- Dut Dutch tions from the vernaculardialects, so at least Eng English the written standard languagesmay have Est Estonian beeninfluenced by Latin, the Europeanwrit- Fin Finnish ten languagepar excellencefor many centu- Fr French ries. Thus, non-standard English has con- Gae ScotsGaelic structionslike I won'tdo nothing('I won't do Grg Georgian anything'), and similarly in non-standard Grk Greek German and French (cf. Haspelmath 199'l: Grm German 205). Analogously, Latin-type relative pro- Hng Hungarian nounsoccur widely in the standardlanguages Ice Icelandic of Europe, but vernacularspeech often pre- Ir lrish fers relative particles (Lehmann 1984: 88, It Iralian ' ^). However, Latin probably only helped Kom Komi t' reinforce these structures in those lan- Lat Latin guageswhere they existed already indepen- Laz Laz dently as variants. Lit Lithunian Thus, we are left with option (iii), the time Ltv Latvian of the great migrations at the transition be- Lzg Lezgian tween antiquity and the Middle Ages. This Mar Mari seemsto be the appropriate time frame at Mlt Maltese 'have'-perfect, least for articles,the the par- Mrd Mordvin ticipial passive,anticausatives, negative in- Nnts Nenets definites, nominative experiencersand verb Nor Norwegian fronting. The rise of these constructions can Pol Polish be observedonly with diffrculty becausethey Prt Portuguese wereby and largeabsent in the written classi- Rorn Romanian cal languagesbut seem to be well in place Rus Russian once the vernacularlanguages appear in the SAE StandardAverage European written record toward the end of the first Sam Saami millennium CE (cf. also Fehling 1980).This SCr Serbian/Croatian hypothesisderives some further plausibility Sln Slovene from the fact that language contact must Spn Spanish have been particularly intensive and effective Srd Sardinian during the great migrations, and in the case Swd Swedish lrench and northern Italian we have am- Tat Tatar ple recordsof the lexicaleffects of thesecon- Trk Turkish tactsfHowever, it is not so easyto fit features Ubv Ubykh su'a'|fas particle comparatives,,,A and-B" Udm Udmurt conjunction and relative pronouns into this Ukr Ukrainian picture, becausethese features seem to have Wel Welsh developedaround the middle of the first mil- lenium BC or even earlier (cf. Haspelmath 1998). Of course, we must always reckon T. References "Preterite with the possibility(or even likelihood) that Abraham, Werner. 1999. decay as a different SAE features are due to different European areal phenomenon". .Folla Linguistica historicalcircumstances, and the correct pic- 33.1:11-18. "Body ture is likely to be much more complicated Ameka, Felix. 1996. parts in Ewe gram- than we can imagineat the moment, let alone mar". In: Chappell.Hilary & McGregor.William discussin this article. (eds.).Ile grannnr oJ inalienabilitt,:A tt'pological I 508 XIV. Typologicalcharacterization of languagelamilies and li'guistic areas

perspective "The on bod,-part ternl.lcnd the part-x.lnle Fehling. Detlev. 1980. origins of European relution.Berlin: Mouton de Cruyter, 78J-8,10. syntax". Folia LinguisticaHktorica, l: 353-387. Bechert, Johannes& Bernini,Giuliano & Buridant. Feuillet.Jack (ed.). 1998.Actance et valencedans Claude (eds.). 1990.Totard a t1,polog1,of European les longues d'Europe. (Empirical Approaches to languages.(Empirical Approachesto LanguageTy- Language Typology-EUROTYR 20-2.) Berlin: pology, 8.) Berlin: Mouton de Gruyter. Mouton de Gruyter. Beckman, Natanael. "Standard 1934. Vtisteuropeisksvntox; Garvin, Paul. 1949. AverageEuropean Nagru nvbildningur i tnrdiska och undt.u rii.staurtt- and Czech".Studia Linguistica 3: 65-85. peiska sprdk. [West Europeansyntax: Some inno- Gensler, Orin D. 1993.A typological vatlve constructionsin the Nordic and other West evaluationof CelticlHanito-Senitic syntactic parallels. European languages.]G

sodicsystems of Europeanlanguages". In: van der sa,-s it1tnennrv oJ Edx'ard S-apir.Menasha, Wis.: Hulst. Harry (ed.) ll/ordprosodic s1'stenrs in the lan- Sapir Memorial PublicationFund. 75-93. [Re- guagesof Europe.(Empirical Approachesto Lan- printedin Whorf (1956),134- 159.1 guageTypology-EUROTYP. 20-4.) Berlin:Mou- Whorf, Benjamin Lee. 1956. Language,thought, ton de Gruyter,425-47 5. ond realitl': Selected +tritirtgs of Beniatnin Lee Vennemann,Theo. 1994."Linguistic reconstruction ll4nrl'. Edited by John B. Carroll, Cambridge/MA: in the context of European prehistory". Ti'ansac- MIT Press. tionsof the PhilologicalSociety 92:213-282. Wintschalek.Walter. 1993.Die Areallinguistikanr Wagner,Heinrich. 1959.Dns Verbunritr den Spra- Beispiel$,nt aktischer Abereinstinnrungen in llblga- chender briti.schenInseln: ein Beiftag:ur geographi- Kann-Areal. (Studia Uralica, 7.) Wiesbaden:Har- schenTypologie des Verbunrs.Ttibingen: Niemeyer. rassowitz. "The Whorf, BenjaminLee. 1941. relationof habit- ual thoughtand behaviorto language".In: Spier. Martin Haspelmath, MPI Leipzig Leslie(ed.) Language,t'ulture, and personality: Es- ( Deutschland)

l flj llil i

t 108.Aire linguistiquebalkanique t. r l. G6n6ralites effet, les traits communssont trop nombreux 2. Phonologie pour qu'ils soient le fruit du hasard. ll est t. 3. Systdmeverbal vrai que les sp6cialistesdiscutent encorede 4. Systdmenominal la notion de < balkanisme), que I'on dehnira 5. Autresunites ici comme un trait typologique propre ir au 6. Relationsphrastiques T. Subordination moins trois languesde I'union. Ce trait n'a ti 8. R6l-6rences pas besoind'€tre unique en son genre(ainsi, t:,ll I'article d6fini postposeexiste dans les lan- guesscandinaves, le < redoublement> de I'ob- 1. G6neralites [,iliil;] jet se retrouve dans les languesromanes); il sl;iri;I La linguistiquebalkanique est une discipline doit 6tre le r6sultatd'une convergencequi relativementr6cente, bien que la ddcouverte aboutit I un r6sultatidentique ou quasi iden- de traitscommuns entre les langues balkani- tique, alors qu'il n'existait pas ir des stades $iliiiir ques remonte ii la premidremoiti6 du XIX" plus anciens. sidcle.Les specialistes(Asenova 19'19:5-45; Les tAchesde la linguistique balkanique Schaller1975: 37-45) s'accordentd diviser sont consign6esdans l'histoire de la disci- I'histoirede la disciplineen trois periodes: pline. Ellesont un triple aspect:synchronique une pdriode prtlinmnire, ou I'on cherche ir (description) panchronique (extension) et expliquer les traits communs par I'influence diachronique (formation et 6volution). Bien du substrat,une p6riode classicpteoi la lin- que I'essentieldu travail descriptif semble guistique balkanique acquiert ses lettres de avoir et6 acompli (la monographiede Sand- noblessegrdce 2r la publication en 1930 de feld a 6te compl6tee,souvent am6lioree,par Linguistique balkanique. Problimes et rtsul- des centainesd'articles et d'6tudesde d6tail tats de Sandfeld,qui repr6sentela premiere qui ont permis d'accroitre et d'approfondir synthdsecompldte, et une periode modente, les donn6es),il reste toujours beaucoup ir marqueepar le polycentrismeet I'internatio- faire. L 6tude de I'extensiondes balkanismes nalisationdes recherches(nombreuses revues n6cessitele recoursir la geographielinguisti- specifiqueset organisationde congres). que (ou linguistiqueareale) pour determiner La linguistiquebalkanique ne cottsistepas avecexactitude le lieu d'apparition de chaque ir juxtaposerdes descriptionsde languesdi- balkanismeet son extensionr6elle sur le ter- versesdont le seul lien serait la contiguit6 rain. Enfin, la perspectivediachronique n'est g6ographique:il faut que ces langues for- jamais perdue de vue par les balkanologues, ment une < union linguistique> (Sprachbundl. malgr6 les nombreusesdifficult6s auxquelles MCmesi certainesvoix s'6ldventencore pour ils sont confront6s,faute de documents6crits. nier la r6alit6de I'union balkanique(Andrio- Trois aspectssont ir prendre en consid6ra- tis & Kourmoulis 1968),la plupart des lin- tion: l) La gendsede I'union linguistique guistessont convaincusde son existence.En balkanique;2) La genesedes balkanismes;3)