The European linguistic area: Standard Average European (2001)

1492 XIV. Typological characterization of language families and linguistic areas

107. The European linguistic area: Standard Average European

1. Introduction2. The major SAE features3. Some further likely SAE features4. Degrees of membership in SAE5. How did SAE come into being?6. Abbreviations of language names7. References

1. Introduction

This article summarizes some of the mainpieces of evidence for a linguistic area (orSprachbund) in Europe that comprises theRomance, Germanic and Balto-Slavic lan-guages, the Balkan languages, and more mar-ginally also the westernmost Finno-Ugrianlanguages (these will be called core Europeanlanguages in this article). This linguistic areais sometimes called Standard Average Euro-pean (abbreviated SAE), following Whorf(1941) [1956: 138]. The existence of this lin-guistic area is a relatively new insight (cf.Bechert et al. 1990, Bernini & Ramat 1996,Haspelmath 1998, van der Auwera 1998, Kö-nig & Haspelmath 1999).

While the close syntactic parallels amongthe Balkan languages have struck linguistssince the 19th century and the existence ofa Balkan Sprachbund has been universallyaccepted, the European linguistic area haslong been overlooked. This may at first ap-pear surprising, because the members of theSprachbund are among the best studied lan-guages of the world. However, it is easy tounderstand why linguists have been slow toappreciate the significance of the similaritiesamong the core European languages: Sincemost comparative linguists know these lan-guages particularly well, they have tended tosee non-European languages as special andunusual, and the similarities among theEuropean languages have not seemed sur-prising. Thus, it was only toward the end ofthe 20th century, as more and more had be-come known about the grammatical proper-ties of the languages of the rest of the world,that linguists realized how peculiar the coreEuropean languages are in some ways whenseen in the world-wide context. From thisperspective, Standard Average European mayeven appear as an “exotic language” (Dahl1990).

A linguistic area can be recognized whena number of geographically contiguous lan-

guages share structural features which cannotbe due to retention from a common proto-language and which give these languages aprofile that makes them stand out among thesurrounding languages. There is thus no min-imum number of languages that a linguisticarea comprises (pace Stolz 2001a). In prin-ciple, there could be a linguistic area con-sisting of just two languages (though thiswould be rather uninteresting), and thereare also very large (continent-sized) linguisticareas (Dryer 1989a). Likewise, there is nominimum number of structural features thatthe languages must share in order to qualifyas a Sprachbund. For instance, Jakobson(1931) establishes his “Eurasian linguisticarea” on the basis of just two phonologicalfeatures, but of course an area that sharesmore features is more interesting. As will beshown below, Standard Average Europeanlanguages share over a dozen highly charac-teristic features, so we are dealing with a veryinteresting Sprachbund.

A linguistic area is particularly strikingwhen it comprises languages from genealog-ically unrelated languages (like the SouthAsian linguistic area (J Art. 109), or theMesoamerican linguistic area (J Art. 110)),but this is not a necessary feature of aSprachbund. The Balkan languages are allIndo-European, but they are from differentfamilies within Indo-European (Romance,Slavic, Greek, Albanian), and not all lan-guages of these families belong to the Balkanlinguistic area, so nobody questions the va-lidity of the Balkan Sprachbund (J Art. 108).In the case of SAE, three entire branchesof Indo-European (Romance, Germanic andBalto-Slavic) belong to the linguistic area.However, here too it is clear that we arenot dealing with a genealogical grouping,because nobody ever proposed a branch ofIndo-European that consists of preciselythese three families. On the contrary, Indo-Europeanists typically assume a particularlyclose genealogical relationship between Italicand Celtic (and sometimes even an Italo-Celtic protolanguage), but Romance (the soledescendant of Italic) is inside SAE, while theCeltic languages do not belong to SAE. Andsince so much is known about the grammat-ical properties that Proto-Indo-Europeanmust have possessed, it is fairly easy to testwhether an SAE feature is an Indo-Euro-

Brought to you by | Max-Planck-Gesellschaft - WIB6417Authenticated | 194.94.96.194

Download Date | 2/18/13 9:31 AM


peanism or not. As was shown in Haspelmath(1998), most of the characteristic SAE fea-tures (also called Europeanisms here) are notIndo-Europeanisms but later common inno-vations.

Thus, what needs to be shown in order todemonstrate that a structural feature is aEuropeanism is

(i) that the great majority of core Europeanlanguages possesses it;

(ii) that the geographically adjacent lan-guages lack it (i. e. Celtic in the west,Turkic, eastern Uralic, Abkhaz-Adygh-ean and Nakh-Daghestanian in the east,and perhaps Afro-Asiatic in the south);

(iii) that the eastern Indo-European lan-guages lack it (Armenian, Iranian, In-dic); and

(iv) that this feature is not found in the ma-jority of the world’s languages.

Particularly the last point is not easy to de-monstrate for many features because thereare still far too few representative world-widestudies of grammatical structures, so to theextent that our knowledge about the world’slanguages is incomplete and biased, we can-not be sure about the European linguisticarea. In this article, I will cite whatever in-formation is available, and sometimes I willhave to resort to impressionistic observa-tions.

The designation “core European lan-guage” for members of SAE is deliberatelyvague, because the European linguistic areadoes not have sharp boundaries. It seemspossible to identify a nucleus consisting ofcontinental West Germanic languages (e. g.Dutch, German) and Gallo-Romance (e. g.French, Occitan, northern Italo-Romance).For this set of languages, van der Auwera(1998a: 824) proposes the name CharlemagneSprachbund. Of the other languages, thosewhich are geographically further from thiscenter also seem to share significantly fewerSAE features, i. e. Ibero-Romance, insularScandinavian (Icelandic and Faroese), EastSlavic (Russian, Ukrainian, Belorussian) andBaltic. Even English, a West Germanic lan-guage, is clearly not within the nucleus. Ofthe non-Indo-European languages of Europe,the western Uralic languages (i. e. Hungarianand Balto-Finnic) are at least marginal mem-bers of Standard Average European; they arein many ways strikingly different from east-ern Uralic. Maltese also exhibits a number ofEuropeanisms not shared by other Arabic

varieties, but Basque seems to show very fewof them. Somewhat further to the east, Geor-gian in the southern Caucasus (and perhapsthe other Kartvelian languages) shares asurprising number of features with the coreEuropean languages. These impressionisticstatements should eventually be quantified,but since it is not clear how much weightshould be attached to each feature, this is notstraightforward.

All of the features discussed below are syn-tactic, or concern the existence of certainmorphosyntactic categories. I am not awareof any phonological properties characteristicof the core European languages (cf. Jakob-son 1931: 182: “do six por ne udalos’ najti niodnogo obsceevropejskogo … polozitel’nogofonologiceskogo priznaka [so far not a singleEurope-wide positive phonological feature hasbeen found]”). Perhaps phonologists havenot looked hard enough, but at least one ma-jor recent study of word prosody in Euro-pean languages has not found any phonolog-ical evidence for Standard Average European(van der Hulst et al. 1999, especially Maps1�4) (but cf. Pisani 1969). A few generaliza-tions are discussed by Ternes (1998), but hefinds that in most respects European lan-guages are unremarkable from a world-wideperspective. Perhaps the only features worthmentioning are the relatively large vowel in-ventories (no 3-vowel or 4-vowel inventories)and the relatively common consonant clus-ters (no restriction to CV syllables). In theserespects, European languages are not average,but they are by no means extreme either.

2. The major Standard AverageEuropean features

In this section I will discuss a dozen gram-matical features that are characteristic of thecore European languages and that togetherdefine the SAE Sprachbund. In each case Iwill briefly define the feature and give a fewexamples from SAE languages. Then a namemap, which indicates the approximate loca-tion of languages by the arrangement of (ab-breviated) language names, shows the distri-bution of the various feature values withinEurope. In each case it can be observed thatthe nuclear SAE languages are within theSAE isogloss, and that the marginal lan-guages tend to be outside the isogloss to agreater or lesser extent. (Part of the materialpresented here was already included in Has-pelmath 1998.)




2.1. Definite and indefinite articlesBoth a definite and an indefinite article (e. g.English the book/a book; J Art. 62) exist inall Romance and almost all Germanic lan-guages plus some of the Balkan languages(Modern Greek, perhaps Albanian and Bul-garian), but not outside Standard AverageEuropean. To be sure, their forms and syn-tactic behavior show considerable diversity(see Nocentini 1996 for an overview), buttheir very existence is characteristic enough.The distribution of articles in European lan-guages is shown in Map 107.1. (Abbrevi-ations of language names are given in the Ap-pendix.)

-------- definite and indefinite article present- - - - only definite article present

Map 107.1: Definite and indefinite article

In large parts of eastern Europe there areno articles at all (East Slavic, West Slavic,Finno-Ugrian other than Hungarian, Turkic,Nakh-Daghestanian, Kartvelian). Some neigh-boring non-SAE languages do have definitearticles (e. g. Celtic, Semitic, Abkhaz, Mord-vin), and Turkish has an indefinite article,but no neighboring non-SAE language hasboth definite and indefinite articles. The onlyexception among Germanic languages, Ice-landic (which only has definite articles likenearby Celtic), is also the most peripheralGermanic language geographically. We canalso be certain that the existence of definiteand indefinite articles is not an Indo-Euro-peanism: The Iranian and Indic languageshave generally lacked articles throughouttheir history.

World-wide, articles are not nearly ascommon as in Europe: According to Dryer’s(1989b: 85) findings, “it appears that about athird of the languages of the world employarticles” (125 out of a sample of about 400languages). Only 31 languages of those inDryer’s sample (i. e. less than 8%) have bothdefinite and indefinite articles.

2.2. Relative clauses with relative pronounsThe type of relative clause found in languagessuch as German, French or Russian seems tobe unique to Standard Average Europeanlanguages. It is characterized by the follow-ing four features: The relative clause is post-nominal, there is an inflecting relative pro-noun, this pronoun introduces the relativeclause, and the relative pronoun functions asa resumptive, i. e. it signals the head’s rolewithin the relative clause (cf. Lehmann 1984:103�109, Comrie 1998). In English, a rela-tive construction like the suspicious womanwhom I described also displays all these fea-tures. Furthermore, in most SAE languagesthe relative pronoun is based on an interrog-ative pronoun (this is true of all Romance, allSlavic and some Germanic languages, Mod-ern Greek, as well as Hungarian and Geor-gian). (Languages like German, whose rela-tive pronoun is based on a demonstrative, orFinnish, which has a special relative pro-noun, are not common.) The geographicaldistribution of the relative pronoun strategyis shown in Map 107.2.

-------- relative clause with introducing relative pro-noun

- - - - only particle relative clause

Map 107.2: Two relative clause types in Europe

The only other type that is widespread inEurope is the postnominal relative clauseintroduced by a relative particle (Lehmann1984: 85�87), which often occurs in the samelanguage beside the resumptive relative pro-noun type just described (an English examplewould be the radio that I bought). Particlerelatives of this type exist in most Slavic andRomance languages, as well as in Scandina-vian languages and Modern Greek, but alsoin Welsh and Irish (Lehmann 1984: 88�90).The relative particle is sometimes difficult todistinguish from a degenerate resumptivepronoun, and in many European languages




it developed from a relative pronoun throughthe gradual loss of inflectional distinctions.However, this also means that the relativeclause loses its specifically European flavor,because particle relatives are also attestedwidely elsewhere in the world (e. g. in Per-sian, Modern Hebrew, Nahuatl, Indonesian,Yoruba, and Thai, cf. Lehmann 1984: 85�97).

However, the relative pronoun strategyclearly is typically European. It is not foundin the eastern Indo-European languages, andas Comrie (1998: 61) notes, “relative clausesformed using the relative pronoun strategyare quite exceptional outside Europe, exceptas a recent result of the influence of Euro-pean languages … The relative pronounstrategy thus seems to be a remarkable arealtypological feature of European languages,especially the standard written languages”.

2.3. ‘Have’-perfectAnother well-known feature typical of SAElanguages is the (transitive) perfect formed by‘have’ plus a passive participle (e. g. EnglishI have written, Swedish jag har skrivit, Span-ish he escrito; J Art. 59). A perfect of thiskind exists in all Romance and Germanic lan-guages plus some of the Balkan languages(Albanian, Modern Greek, Macedonian), andalso in Czech (Garvin 1949: 84). These per-fects do not all mean the same thing, becausethey are at different stages in the grammati-calization process: in French and German,the perfect can be used as a normal perfectivepast, including the function of a narrativetense, while in Spanish, English and Swedishthe perfect has a distinct present-anteriormeaning. What is important here is that theyall must have had basically the same meaningwhen they were first created. The geographi-cal distribution of ‘have’-perfects in Europeis shown in Map 107.3.

Map 107.3: ‘Have’-perfects in Europe

In contrast to the languages just mentioned,in Slavic, Finno-Ugrian and Armenian theperfect is usually based on a participialconstruction with an active participle and acopula (e. g. Finnish ole-n saa-nut [be-1sg

receive-ptcp] ‘I have received’). Hungarianseems to lack a perfect completely. In someNakh-Daghestanian languages (e. g. Lezgianand Godoberi), the perfect is formed on thebasis of the past converb plus the copula.Georgian comes closest to the SAE prototypein that its transitive perfect is based on a pas-sive participle, but this is combined with thecopula rather than the transitive verb ‘have’,so that the perfect has a quasi-passive struc-ture, with the agent in the dative case (‘Theletter is-written to-me’, rather than ‘I have-written the letter’). In Welsh, the perfect isformed with the preposition wedi ‘after’ (‘Sheis after selling the house’ for ‘She has soldthe house’). The eastern Indo-European lan-guages also lack a ‘have’-perfect (for in-stance, both Persian and Hindi/Urdu have aperfect based on a participle plus the copula,somewhat like Slavic and Armenian).

Dahl (1995, 1996: 365), taking a globalperspective, notes that the ‘have’-perfect is al-most exclusively found in Europe. Now onemight object that this is not a primitive fea-ture of European languages. Many languagesdo not use a transitive ‘have’-verb for indi-cating predicative possession at all, and it hasin fact been suggested that the very existenceof a transitive verb of predicative possessionis a Europeanism (e. g. Lazard 1990: 246�47;Benveniste 1960 [1966: 195]: “L’expression laplus courante du rapport indique dans noslangues par avoir s’enonce a l’inverse par etrea … Telle est la situation dans la majorite deslangues.”) The restriction of a ‘have’-perfectto Europe would then be just a consequenceof this (cf. Dahl 1990: 7). However, so far nopublished research has documented an arealrestriction for ‘have’ verbs. From Heine’s(1997: 47�50, 240�44) survey of predicativepossessive constructions, not much supportcan be drawn for such a claim. Still, this isan interesting idea to be addressed by furtherresearch. If ‘have’-verbs turn out to be typi-cal of Europe, that would fit with the ten-dency of European languages to have nomi-native experiencers in experiential verbs (seethe next section).

2.4. Nominative experiencersThere are two ways of expressing experiencerarguments of verbs of sensation, emotion,cognition and perception: The experiencer




may be assimilated to agents and coded asa nominative subject (e. g. I like it), or it maybe assimilated to a patient or goal, so thatthe stimulus argument is coded as the nomi-native subject (e. g. It pleases me). In Bos-song’s (1998) typology, the first type is calledgeneralizing, and the second type is calledinverting. Bossong studies the expression often common experiential predicates in 40European languages. He computes the rela-tion between inverting predicates and gener-alizing predicates, arriving at figures between0.0 for English (where all predicates aregeneralizing) and 5.0 for Lezgian (where allpredicates are inverting). By arbitrarily divid-ing the languages into those showing pre-dominant generalization (ratios between 0.0and 0.8) and those showing predominantinversion (ratios between 0.8 and 5.0), wearrive at the geographical pattern shown inMap 107.4.

Map 107.4: Predominant generalization (center) vs. inversion (periphery)

Thus, Bossong’s study basically confirmsearlier claims (Lazard 1990: 246�47, Dahl1990: 7) that the generalizing type is charac-teristic of SAE, although some of the fig-ures are perhaps a bit surprising (e. g. the factthat Hungarian turns out to be more SAEthan German or Dutch, and the inclusion ofTurkish, but not Romanian or Albanian,with respect to this feature). It is not possibleto explain everything here, but we evidentlyhave before us a fairly typical SAE patternwith French and English at the center, Celtic(plus Icelandic this time) at the western mar-gin, Balto-Slavic, Finno-Ugrian and Cauca-sian at the eastern margin, and fairly gradualtransitions within the macro-areas. No sys-tematic world-wide studies have been made,but at least the behavior of eastern Indo-

European is fairly clear: Indic languages arewell-known for their “dative subjects” ofexperiencer verbs, so again the feature isnot genetic (see also Masica 1976, especiallyMap 6, for the areal distribution of dativesubjects in Eurasia and northern Africa).(See Haspelmath 2001 for more discussionof experiential predicates in European lan-guages.)

2.5. Participial passiveStandard Average European languages typi-cally have a canonical passive construction(J Art. 67) formed with a passive participleplus an intransitive copula-like verb (‘be’,‘become’, or the like). In this passive theoriginal direct object becomes the subject andthe original subject may be omitted, but itmay also be expressed as an adverbial agentphrase. Such constructions occur in all Ro-mance and Germanic languages, but also in

all Slavic (including East Slavic) and Balkanlanguages, as well as in Irish. The geographi-cal distribution of such participial passives isshown in Map 107.5.

Map 107.5: Participial passives in Europe




No passives exist in Nakh-Daghestanian andin Hungarian, and passives of different for-mal types are found in Turkic, Georgian, andArmenian (stem suffix), in Basque, and inCeltic (cf. the Welsh ‘get’-passive: ‘Terry gothis hitting by a snowball’ for ‘Terry got hitby a snowball’). Finnish and Irish have pas-sives of a different syntactic type: In this con-struction, only the subject is backgrounded,while the direct object remains in its place.

Participial passives are very rare in lan-guages other than Standard Average Euro-pean. In Haspelmath (1990) I surveyed aworld-wide sample of eighty languages andfound that a passive exists only in the mi-nority of the languages (thirty-one). Of thesethirty-one languages, only four have a pas-sive formed from a participle plus an intran-sitive auxiliary, and two of them are Euro-pean languages (Latin and Danish). Themost common formal type of passive is thestem suffix (found in twenty-five languages).Syntactically, the possibility of an adverbialagent phrase is also by no means universal,but it is characteristic of SAE languages (La-zard 1990: 246).

It must be admitted that the SAE status ofthis feature is less evident than that of thefirst two features because the eastern Indo-European languages also tend to have pas-sives of this type. In fact, in my 1990 study,the two non-European languages with parti-ciple-auxiliary passives were Baluchi (an Ira-nian language) and Maithili (an Indic lan-guage). Thus, one might say that this featureis an Indo-European genealogical feature.However, at least the Celtic languages andArmenian, two non-SAE branches of Indo-European, do not have such passives, andMaltese is a non-Indo-European languagewith such a passive (calqued from Italian).

2.6. Anticausative prominenceThere are three ways in which languages canexpress inchoative-causative alternations suchas ‘get lost/lose’, ‘break (intr.)/break (tr.)’,‘rise/raise’. One is by means of a causativederivation (J Art. 66), i. e. a derived verbbased on the inchoative member of the al-ternation, e. g. Mongolian xajl-uul- ‘melt (tr.)’,from xajl- ‘melt (intr.)’. The second is bymeans of an anticausative derivation, i. e. aderived verb based on the causative member,e. g. Russian izmenit’-sja ‘change (intr.)’, fromizmenit’ ‘change (tr.)’. (The third type, inwhich neither member is derived from theother, i. e. non-directed alternations, will not

be considered further here.) In Haspelmath(1993), I examined 31 verb pairs in 21 lan-guages and found that languages differgreatly in the way inchoative-causative pairsare expressed: Some languages are anticau-sative-prominent, preferring anticausatives tocausatives, while others are causative-promi-nent. It turns out that anticausative-promi-nence is a characteristic feature of SAE. Inmy sample, German, French, Romanian,Russian, Modern Greek and Lithuanianshow the highest percentages of anticausativeverb pairs (between 100% and 74% of allpairs that do not belong to the third, non-directed, type). The percentage in the Euro-pean languages of my sample are shown inMap 107.6.

-------- 70�100 % anticausatives- - - - 50�70 % anticausatives

Map 107.6: Percentage of anticausative pairs

By contrast, Asian languages show muchlower percentages of anticausatives, prefer-ring causatives instead (e. g. Indonesian: 0%,Mongolian: 11%, Turkish: 34%, Hindi/Urdu35%, Lezgian: 40%). An intermediate posi-tion is occupied by the Finno-Ugrian lan-guages of eastern Europe (Finnish 47%,Udmurt 46%, Hungarian 44%) as well asGeorgian (68%) and Armenian (65%). In astudy involving more languages from Asia,Africa and Europe but less language-partic-ular detail, Masica (1976) found a clear dis-tinctive pattern for Europe: few causatives,heavy reliance on anticausatives (see espe-cially his Maps 2 and 3). In a recent world-wide study of 18 verbs from 80 languages,Nichols et al. (to appear) report that in in-choative-causative pairs involving inanimateparticipants (i. e. the most typical subtype),the causative is generally favored worldwideand is strongly disfavored only in Europe.




Anticausative-prominence is not an Indo-Europeanism: Older Indo-European had aproductive causative formation, which lostits productivity in the European branches,but continued to be productive in easternIndo-European (cf. the low figure of 35% an-ticausatives in Hindi/Urdu).

2.7. Dative external possessorsIn König & Haspelmath (1998) and Haspel-math (1999), we studied the distribution ofexternal possessors in thirty European lan-guages (J Art. 73). We found three main lan-guage types in Europe: (i) those with dativeexternal possessors, e. g. German Die Mutterwäscht dem Kind die Haare ‘The mother iswashing the child’s hair’, (ii) those with loca-tive external possessors, e. g. Swedish Nagonbröt armen pa honom ‘Someone broke hisarm (lit. on him)’, and (iii) those that lackexternal possessors and must express posses-sors NP-internally, e. g. English. The SAEfeature, external possessors in the dative, isfound in Romance, Continental West Ger-manic, Balto-Slavic, Hungarian and Balkanlanguages (Greek, Albanian). North Ger-manic and Balto-Finnic languages have loca-tive external possessors, i. e. they are some-what peripheral SAE languages with respectto this feature. The geographical distributionis shown in Map 107.7.

Map 107.7: Dative external possessor

In the far west (Welsh, Breton, English) andin the southeast (Turkish, Lezgian) of Europethere are languages which do not have exter-nal possessors at all. The eastern Indo-Euro-pean languages Kurdish, Persian and Hindi/Urdu also belong to this type. Outside Europea fourth type enjoys considerable popularity:the “relation-usurping” type, where he pos-sessor “usurps” the syntactic relation of thepossessum (e. g. Chichewa, a Bantu language,has ‘The hyena ate the hare the fish’ for ‘The

hyena ate the hare’s fish’). This type is notfound in Europe at all. Conversely, dative ex-ternal possessors seem to be very rare outsideEurope (the only case I am aware of is Ewe,cf. Ameka 1996), so this is a very robust ex-ample of an SAE feature.

2.8. Negative pronouns and lack ofverbal negation

The areal distribution of negation in Europehas been studied in detail by Bernini &Ramat (1996) (see also Ramat & Bernini1990). Here I will single out just one aspectof negation, the cooccurrence of verbal nega-tion with negative indefinite pronouns. I dis-tinguish two main types: (i) V � NI (verb �negative indefinite), e. g. German Niemandkommt ‘nobody comes’, and (ii) NV � NI(negated verb � negative indefinite), e. g.Modern Greek Kanenas dhen erxete ‘nobody(lit. not) comes’. A third, mixed type mightbe distinguished in which verbal negationcooccurs with negative indefinites only whenthe indefinite follows the verb but not whenit precedes it, e. g. Italian Nessuno viene ‘no-body comes’, but Non ho visto nessuno ‘Not Ihave seen nobody’. For our purposes we canclassify this type as a subtype of (i), V � NI.

The Standard Average European type isV � NI (cf. Bernini & Ramat 1996: 184, Has-pelmath 1997: 202). It is found in French (ifwe disregard the particle ne), Occitan and allGermanic languages, as well as (in the mixedvariety) in Ibero- and Italo-Romance and Al-banian (but not in Romanian or other Bal-kan languages). The geographical distribu-tion of the types is shown on Map 107.8.

Map 107.8: Languages lacking verbal negationwith a negative indefinite

All the eastern European languages (Balto-Slavic, Finno-Ugrian, Turkic, Nakh-Daghes-tanian) with the exception of Georgian, andthe Celtic languages in the west show theNV � NI type. This type is also that of the




eastern Indo-European languages (Iranianand Indic), as well as that of the clear major-ity of the world’s languages: Kahrel (1996)has studied negation in a representative world-wide sample of 40 languages and found onlyfive languages with V � NI negative pat-terns, one of which is the SAE languageDutch (the other four are Mangarayi (Aus-tralia), Evenki, Chukchi (Siberia), and Nama(southern Africa)), as against 41 NV � NIpatterns, and seven others. I found a very sim-ilar pattern in my (non-representative) sampleof 40 languages (Haspelmath 1997: 202).

2.9. Particles in comparative constructionsComparative constructions were investigatedby Stassen (1985) in a world-wide study of 19languages (J Art. 75). Stassen distinguishessix main ways in which the standard of com-parison may be expressed: Three kinds of loc-ative comparatives (‘bigger from X’, ‘biggerto X’, ‘bigger at X’), the exceed comparative(‘Y is big exceeding X’), the conjoined com-parative (‘Y is big, X is little’), and the par-ticle comparative (‘bigger than X’). The par-ticle in this latter type is often related to arelative pronoun (cf. English than/that, Latinquam/qui), and the case marking of the stan-dard is not influenced by the particle (so thatit is possible to distinguish ‘I love you morethan she’ from ‘I love you more than her’).

As Heine (1994) notes, the six types are notevenly distributed among the languages ofthe world. Of the 18 particle comparativesin Stassen’s sample, 13 are in Europe, and ofthe 17 European languages in the sample, 13have a particle comparative. The distributionwithin Europe again conforms to our expec-tations: Particle comparatives are found inGermanic, Romance, Balto-Slavic, the Bal-kans, Hungarian, Finnish and Basque, so thisis the SAE type. The distribution is shown inMap 107.9.

-------- particle comparative- - - - locative comparative

Map 107.9: Comparative types in Europe

The locative comparatives are all at the west-ern fringe (Breton) or the eastern fringe ofEurope (Finnish, Russian, Nenets, Ubykh,Turkish, Laz). The other two types do notexist at all in Europe � the exceed compara-tive is found particularly in Africa, and theconjoined comparative occurs only in theAmericas and Oceania.

2.10. Relative-based equative constructionsComparison of equality (equative construc-tions) is discussed less often than comparisonof inequality, and nobody has undertaken astudy of equatives on a world-wide scale.Still, there are good reasons to think thatequative constructions provide evidence forStandard Average European (Haspelmath &Buchholz 1998). In Europe, many languageshave an equative construction that is basedon an adverbial relative-clause construction.For example, Catalan has tan Z com X ‘as Zas X’ (where Z is the adjective and X is thestandard). Catalan com is an adverbial rela-tive pronoun, and tan is a correlative demon-strative. A very similar construction is foundelsewhere in Romance (Portuguese tao Zcomo X, Occitan tan Z coma X), in Germanic(German so Z wie X), in Slavic (Czech tak Zjako X, Russian tak(oj) ze Z kak X), in Ro-mani (kade Z sar X), in Hungarian (olyan Zmint X), in Finnish (niin Z kuin X), and inGeorgian (isetive Z rogorc X). In the Englishconstruction, the relative-clause origin of asis not fully transparent synchronically, butdiachronically as derives from a demonstra-tive (eall swa � all so) that was also usedas a relative pronoun. In some Balkan lan-guages, the correlative demonstrative is notused (e. g. Bulgarian xubava kato tebe ‘aspretty as you’), but the standard marker isclearly of relative-pronoun origin. (There isprobably some connection between the rela-tive-pronoun origin of equative markers andthe relative-pronoun origin of comparativestandard markers that we saw in § 2.9.).

Non-SAE languages have quite differentequative constructions. Many SOV languagesin eastern Europe have a special equativestandard marker (Lezgian xiz, Kalmyk sing;also Basque bezain and Maltese daqs), andthe Celtic languages have a special (non-demonstrative) marker on the adjective (e. g.Irish chomh Z le X ‘equative Z with X’). Inthe Scandinavian languages, the word ‘equ-ally’ is used on the adjective (e. g. Swedishlika Z som X ‘equally Z as X’). The distri-bution of the relative-based equative con-




struction in Europe is shown in Map 107.10,following Haspelmath & Buchholz (1998:297).

Map 107.10: Relative-based equative-constructions

Impressionalistically, relative-based equativesseem to be rare in the world’s languages, andthe eastern Indo-European languages do notseem to use them in general (however, acounterexample is Punjabi).

2.11. Subject person affixes as strictagreement markers

The majority of the world’s languages havebound person markers on the verb that cross-refer to the verb’s subject (or agent). Whenthese subject affixes cooccur with overt sub-ject NPs (full NPs or independent subjectpronouns), they are called agreement mark-ers. However, in most languages they can oc-cur on their own and need not cooccur withovert subject NPs. For example, in the Bul-garian phrase vie rabotite ‘you (pl.) work’, wesee the subject suffix -ite (2nd person plural)cooccurring with the independent subjectpronoun vie ‘you (pl.)’, showing that -ite isan agreement marker. But in Bulgarian it isequally possible and probably more commonto say just rabotite ‘you (pl.) work’, i. e. thesubject suffix can have a referential functionon its own. In German, by contrast, this isnot possible: ‘you work’ is ihr arbeit-et. Sincethe agreement suffix -et does not have suchan independent referential function, the sub-ject pronoun ihr cannot be omitted. Lan-guages like German are often called “non-pro-drop languages”, and languages likeBulgarian are called “pro-drop languages”;better terms would be “strict-agreement lan-guages” vs. “referential-agreement languages”.

It has sometimes been thought that strictagreement, as exhibited by German, English,and French, is the norm and that referentialagreement is somehow special. But in fact,referential agreement is far more widespread

in the world’s languages, and strict subjectagreement is characteristic of a few Europeanlanguages, some of which happen to be well-known. In her world-wide sample of 272languages, Siewierska (1999) finds only twostrict-agreement languages, Dutch (an SAElanguage) and Vanimo (a Papuan languageof New Guinea). Siewierska further notesthat outside of Europe, she is aware of onlytwo additional strict-agreement languages thatare not in her sample (Anejom and Labu, twoOceanic languages). Gilligan (1987) reacheda similar conclusion on the basis of a sampleof 100 languages. The distribution of strictsubject agreement markers in some Europeanlanguages is shown in Map 107.11.

-------- languages with strict subject agreement- - - - languages with obligatory subject pronouns,

lacking verb agreement

Map 107.11: Obligatory subject pronouns

The map shows two non-contiguous areas inwhich subject agreement suffixes cannot havea referential function: Germanic and Gallo-Romance languages with Welsh on the onehand, and Russian on the other. Perhaps onlythe western European area should be thoughtof as being relevant for SAE; in Russian,past-tense verbs do not have subject personaffixes, so Russian is not a very good exam-ple of a strict-agreement language. In theeastern Nordic languages (Norwegian, Swed-ish, Danish), the subject pronouns are obliga-tory as they are in English, German or Ice-landic, but the languages have lost agreementdistinctions on the verb entirely (cf. Swedishjag biter/du biter/han biter ‘I/you/he bite(s)’,Icelandic eg bıt/pu bıtur/hann bıtur). Theselanguages are thus “non-pro-drop” in asense, but they are not strict-agreement lan-guages. English is approaching this type, asthe only remnant of subject agreement is the3rd person singular present-tense suffix -s.(There are also some languages of this type




in the eastern Caucasus, and indeed in manyother parts of the world, but they may neverhave had subject person agreement marking.)

2.12. Intensifier-reflexive differentiationIntensifiers are words like English self, Ger-man selbst, French meme and Russian samthat characterize a noun phrase referent ascentral as opposed to an implicit or explicitperiphery (e. g. The Pope himself gave us anaudience, i. e. not just the cardinals (JArt. 57; König & Siemund 1999). In manylanguages, the intensifier expression is alsoused as a reflexive pronoun, for instance inPersian (xod-as ‘himself’: Husang xod-as‘Hushang himself’, and Husang xodas-ra did[Hushang self-acc saw] ‘Hushang saw him-self’). However, a feature that is typical ofSAE languages is the differentiation of reflex-ive pronouns and intensifiers (König & Has-pelmath 1999). For instance, German has sich(reflexive) vs. selbst (intensifier), Russian hassebja vs. sam, Italian has si vs. stesso, Greekhas eafto vs. ıdhjos. Map 107.12 shows the lan-guages in Europe with special reflexive pro-nouns that are not identifical to intensifiers.

Map 107.12: Intensifier-reflexive differentiation

Intensifier-reflexive differentiation is not anIndo-Europeanism, because eastern Indo-European languages have the same expres-sion for intensifiers and reflexives (e. g. Per-sian xod-as, Hindi aap). There are no pub-lished world-wide studies yet, but it seemsthat non-differentiation is very commonaround the world, and while differentiation isalso found elsewhere, it is not found in areasimmediately adjacent to European languages.

3. Some further likely SAE features

In this section, I will mention a few featureswhich are less well-documented than those in§ 2, or whose geographical distribution is less

striking, but which nevertheless seem goodcandidates for Europeanisms. No maps willbe given for these features, and the evidencewill be summarized only briefly.

3.1. Verb fronting in polar interrogativesIn the large majority of languages, polar in-terrogatives are marked by interrogative in-tonation or an interrogative particle or both(J Art. 77). In his sample of 79 languages,Ultan (1978) found only seven languagesshowing the alternative strategy of verbfronting (often called “subject-verb inver-sion”). Of these, six are European (English,French, Romanian, Russian, Hungarian,Finnish; the seventh language is Malay), sothat the SAE status of verb fronting seemsbeyond doubt. In fact, the large majority ofGermanic, Romance and Slavic languages(plus Modern Greek) appear to have verbfronting in polar questions in one form oranother. The three European languages forwhich Ultan explicitly reports that no verbfronting occurs are peripheral: Basque, Gae-lic and Lithuanian. Furthermore, SAE lan-guages are characterized by the absence of aninterrogative particle. In Ultan’s data, thenine European languages exhibiting a par-ticle in polar questions are all peripheral toa greater or lesser extent: Basque, Irish, Scot-tish Gaelic, Albanian, Hungarian, Lithua-nian, Russian, Finnish, Turkish (and I canadd Nakh-Daghestanian). Verb fronting inpolar questions was suggested as a Euro-peanism already by Beckman (1934) (cf.Dahl 1990).

3.2. Comparative marking of adjectivesMost European languages have special formsfor adjectives occurring in comparative con-structions. For instance, English uses thesuffix -er in this way (The dog is bigg-er thanthe cat). Such an inflectional marker of adjec-tives is not common in the world’s languagesoutside of Europe. Some languages use somekind of adverbial particle modifying the ad-jective (‘more’), but perhaps the most com-mon type is represented by Japanese, wherethe comparative semantics is carried by thestandard marker alone (e. g. inu-ga neko yoriookii [dog-subj cat from big] ‘the dog is big-ger than the cat’).

Special comparative forms are found in allGermanic, Balto-Slavic and Balkan lan-guages (with the exception of Romanian andAlbanian), and most Romance languagespreserve at least four suppletive forms (e. g.




Italian maggiore ‘bigger’, minore ‘smaller’,peggiore ‘worse’, migliore ‘better’). Compara-tive forms also exist in Basque (e. g. haundi-ago ‘bigg-er’), Hungarian (nagy-obb ‘bigg-er’),Finnish (iso-mpi ‘bigg-er’), and other Finno-Ugrian languages.

Comparative forms are not completely un-known outside of Europe. Arabic has a spe-cial comparative form (e. g. ?akbar ‘bigger’,from kabiir ‘big’), but it is unique amongAfro-Asiatic languages in this respect. OldIndo-Iranian languages had comparativeforms, and the modern Iranian languageshave preserved them to some extent (e. g. Per-sian -tær, Zaza -er). But further east, in mod-ern Indic, the comparative does not existanymore, and languages like Hindi-Urdu andBengali use a construction analogous to theJapanese example just cited. Similarly, in theUralic languages, the further east we go, thefewer comparatives we find. For instance,Khanty (a Finno-Ugrian language spoken inwestern Siberia, i. e. outside of Europe) doeshave a comparative form in -sek (e. g. jam-sek‘better’), which is used when no standard ispresent. But in a complete comparative con-struction, no marking is found on the adjec-tive (e. g. nan ke:se:-n e:welt jam [you knife-2sg from good] ‘better than your knife’, Ni-kolaeva 1999: 21).

Thus, although this feature is not confinedto Europe, it is typical of a SAE feature inthat it is robustly present in western Indo-European and Uralic languages, but getsrarer the further east we go in these families.

3.3. “A and-B” conjunctionThe feature discussed in this section is lessdistinctive than the others mentioned so far,but I hope to show that it is not at all devoidof interest. Stassen (2000) offers the firstworld-wide typological study of NP conjunc-tion strategies, based on a sample of 260languages (J Art. 82). He distinguishes twobasic types, and-languages (using a symmet-ric particle) and with-languages (using anasymmetric comitative marker). Two thirdsof Stassen’s sample languages are and-lan-guages, and since SAE clearly belongs to thistype, too, it is not a very distinctive property.And-languages cover all of northern Eurasia,South Asia, the Middle East and northernAfrica, Australia, New Guinea, and parts ofCentral and South America. With-languagesare encountered in sub-Saharan Africa, Eastand Southeast Asia, the islands of Oceania,and large areas of North and South America.

However, within the and-languages there areseveral sub-types according to the position ofthe particle, which we may call “A and-B”,“A-and B”, “A-and B-and”, and “A B-and”(of the remaining logical possibilities, “and-AB” seems to be inexistent, and “and-A and-B” occurs only as a secondary pattern). MostEuropean languages, and in particular allSAE languages, belong to the sub-type “Aand-B”. The types “A-and B-and” and “A-and B” are found in some languages of theCaucasus and in some Turkic languages, aswell as scattered throughout northern Eu-rasia and South Asia (e. g. in Abkhaz, Archi,Persian, Sinhalese, Tamil, Burmese, Koreanaccording to Stassen; Stassen also points outthat there is a correlation with verb-finalword order here). Furthermore, some periph-eral European languages make restricted useof the with-strategy (e. g. Russian my s toboj‘I and you’, lit. ‘we with you’, and also OldIrish, Lithuanian, Polish and Hungarian,according to Stassen). Taken together, thesedata do show that belonging to the “A and-B” type is not a trivial feature of the SAElinguistic area.

3.4. Comitative-instrumental syncretismIn all SAE languages, the preposition thatexpresses accompaniment (� comitative) alsoserves to express the instrument role (e. g.English with: with her husband/with the ham-mer). Such languages are said to exhibit com-itative-instrumental syncretism. Stolz (1996)studied comitative and instrumental markersin a world-wide sample of 323 languages andfound that this kind of syncretism is typicalof Europe. Non-European languages morecommonly possess separate markers for thesetwo semantic roles (e. g. Swahili na ‘with(comitative)’, kwa ‘with (instrumental)’. AsTable 107.1 shows, about two thirds of Stolz’ssample languages are non-syncretic, and onlyone quarter is syncretic. (The remaining lan-guages belong to a mixed type, which I ig-nore here for the sake of simplicity; thus, thepercentages do not add up to 100%.)

Two areas diverge significantly from thegeneral trend: Oceania has far less syncretismthan the world average, and Europe has farmore syncretism than the world average.When we look at the pattern within Europe,it becomes even clearer that we are dealingwith an SAE feature (as Stolz recognizes, cf.1996: 120). Of the 16 non-syncretic languagesin Europe, 10 are Caucasian languages, i. e.they are clearly outside of SAE, and one is




Table 107.1: Comitative-instrumental: Syncretic and non-syncretic languages

syncretic (e. g. English) non-syncretic (e. g. Swahili)languages percentage languages percentage

Europe 25 49% 16 31%Africa 20 31% 38 58%Americas 16 21% 54 69%Asia 12 18% 47 71%Oceania 6 10% 54 86%World 79 24% 209 65%

only politically, not anthropologically, inEurope (Greenlandic). Four of the remainingfive languages are also otherwise not typicalinstances of SAE (Basque, Finnish, Maltese,Mari). And when we look at the 38 Indo-European languages in Stolz’s sample, wesee that syncretism cannot be regarded as anIndo-Europeanism: Of the eight Indo-Euro-pean languages not spoken in Europe, onlythree show syncretism, while five show non-syncretism. Thus, in Asia Indo-European lan-guages behave like Asian languages, and thereis no general pattern for Indo-European.

3.5. Suppletive second ordinalMost languages have a suppletive form of theordinal numeral ‘first’, i. e. a form not de-rived from the cardinal numeral ‘one’. Anexample is German, where ‘1st’ is erster (un-related to eins ‘1’), contrasting with otherordinals such as zweiter ‘2nd’ (cf. zwei ‘2’),vierter ‘4th’ (cf. vier ‘4’), and so on. In Stolz’s(2001b) study of 100 languages world-wide,there are 95 languages with special ordinalnumerals, and of these, 78 have a suppletiveword for ‘first’. Thus, languages that say(literally) ‘oneth’ for ‘1st’ are not common.However, the same sample has only 22 lan-guages in which the word for ‘2nd’, too, issuppletive and not derived from ‘2’ (e. g.English second). Thus, most languages have(literally) ‘twoth’ for ‘2nd’. The 22 languagesthat have a suppletive ‘2nd’ word are heavilyconcentrated in Europe: 17 are Europeanlanguages, and this type is clearly the major-ity within Europe (which is represented by 27languages in Stolz’s sample). Of the 10 Euro-pean languages that do not have a suppletivesecond ordinal, six are clearly outside SAE(Basque, Turkish, Armenian, Georgian, Lez-gian, Greenlandic). Among SAE languages,only some Balkan languages (Romanian, Al-banian, Romani) and German lack a supple-tive second ordinal.

This is clearly a very marginal feature ingrammar, but it is intriguing that it shouldshow such a clear geographical distribution.

3.6. Some other characteristics of SAEThe features examined so far present themost striking evidence for Standard AverageEuropean, but there are probably many morefeatures that will turn out to be characteristicof the core European languages in one wayor another. In this subsection, several suchcandidates will be mentioned briefly. The firstfew features in the following list are purelynegative: At first glance, this may seem odd,but of course the lack of a category that iswidespread elsewhere is no less significantthan the presence of a category that is rareelsewhere.

(i) Lack of an alienable/inalienable opposi-tion in adnominal possession (J Art. 72). InNichols’s (1992) world-wide sample, almosthalf of the languages show such an opposi-tion, but no European language does (1992:123). More generally, this opposition is rarerin the Old World and common in the NewWorld, but in Europe it is even less commonthan in Africa and Asia.(ii) Lack of an inclusive/exclusive oppositionin first person non-singular pronouns. Again,this opposition is commonest in the NewWorld and in the Pacific region, but inEurope it is even rarer than in Africa andAsia, as was shown by Nichols (1992: 123).(iii) Lack of reduplicating constructions. Ihave no systematic evidence to back up theclaim that this is a characteristic feature ofEuropean languages, but reduplication is socommon across languages that its almost to-tal absence in the core European languagesbecomes striking. (Interestingly, reduplicationexisted in older Indo-European languages atleast in one construction, the perfect, buteven here it was lost entirely by the MiddleAges.)




(iv) Discourse pragmatic notions such astopic and focus are expressed primarily bysentence stress and word order differences(Lazard 1998: 116). Only the Celtic languagesand French give a very prominent role toclefting, and particles marking discoursepragmatic notions are virtually unknown.(v) SVO basic word order at the level of theclause. This feature is of course found else-where in the world, but in Europe it corre-lates particularly well with the other SAEfeatures. The Celtic languages in the westhave VSO order (except for Breton, which isalso otherwise more SAE than Irish andWelsh), and the eastern languages have SOVword order. Interestingly, Balto-Finnic (Fin-nish, Estonian, etc.) and (less unequivocally)Hungarian have SVO word order, whereasthe eastern Uralic languages have SOV. Simi-larly, the eastern Indo-European languagestend to show SOV word order. (See Dryer1998 for more on word order in the lan-guages of Europe.)(vi) European languages tend to have justone converb (J Art. 83) (cf. Nedjalkov 1998).For instance, Romance languages have thegerundio/gerondif, English has the -ing-form,and Slavic and Balkan languages have theiradverbial participle. The Celtic languages inthe west completely lack such a form, and thelanguages east of SAE tend to have morethan one converb. Otherwise the core Euro-pean languages tend to have adverbial con-junctions (J Art. 63) to make adverbialclauses. According to Kortmann (1997: 344),they have “a large, semantically highly dif-ferentiated inventory of free adverbial sub-ordinators placed in clause-initial position”.More generally, they tend to have finite ratherthan non-finite subordinate strategies (JArt. 100), though a multi-purpose infinitiveusually exists (except for the Balkan lan-guages).(vii) European languages usually have a spe-cial construction for negative coordination,e. g. English neither A nor B, Italian ne A neB, Russian ni A ni B, Dutch noch A noch B,Hungarian sem A sem B. Again, no world-wide study has been published, but such anegative coordinating construction is rarelyreported from languages outside Europe (cf.Haspelmath to appear).(viii) SAE languages have a large number ofcharacteristic properties in the area of phasaladverbials (expressions like already, still, nolonger, not yet) (van der Auwera 1998b).These are rather well documented, but for the

detail I have to refer the reader to van derAuwera’s thorough study.(ix) “Preterite decay”: the loss of the oldpreterite and its replacement by the formerpresent perfect. This is a change that oc-curred in the last millenium in French, Ger-man and northern Italian, as well as in someother adjacent European languages (cf. Thie-roff 2000: 285). Its distribution is far nar-rower than that of the other Europeanisms,but it is the only feature of those studied byThieroff whose geography comes close toStandard Average European (cf. also Abra-ham 1999).

Quite a few additional features have beenmentioned in the earlier literature as charac-teristic of SAE, but earlier authors havesometimes neglected to make sure that a pro-posed Europeanism is not also common else-where in the world. Most of Whorf’s originalexamples of SAE features seem to be of thiskind. For instance, he notes that in contrastto SAE, Hopi lacks “imaginary plurals” (suchas ‘ten days’, according to Whorf a “meta-phorical aggregate”). But of course, we haveno evidence that such plurals of time-spannouns are in any way characteristic of Euro-pean languages. It may well be that they arecommon throughout the world. (To giveWhorf his due, it must be added that he wasnot interested in demonstrating that SAElanguages form a Sprachbund. He just usedthis term as a convenient abbreviation for“English and other European languageslikely to be known to the reader”, withoutnecessarily implying that these languages arean exclusive club.)

4. Degrees of membership in SAE

Membership in a Sprachbund is typically amatter of degree. Usually there is a core oflanguages that clearly belong to the Sprach-bund, and a periphery of surrounding lan-guages that share features of the linguisticarea to a greater or lesser extent.

In order to quantify the degrees of mem-bership in SAE, a simple procedure suggestsitself that was first applied to areal typologyby van der Auwera (1998a). In addition toindividual maps in which the lines denote iso-glosses (as in Maps 107.1�12), we can com-bine different features in a single map andshow the number of isoglosses shared by thelanguage. Map 107.13 shows such a “clustermap” in which the lines stand for “quantified




isoglosses” (or “isopleths”). The map com-bines nine features of § 2.: definite and indefi-nite articles, relative clauses with relative pro-nouns, ‘have’-perfect, participial passive, da-tive external possessors, negative pronounsand lack of verbal negation, relative-basedequative constructions, subject person affixesas strict agreement markers, and intensifier-reflexive differentiation. The languages in thenucleus (French and German) show the SAEvalue in all nine of these features. The lan-guages in the next layer (Dutch, other Ro-mance, Albanian) show eight features, thenext layer (English, Greek, Romanian) showsseven features, and so on. In this map, theresulting picture is actually very clear, be-cause the SAE area with at least five SAEfeatures stands out from the remaining lan-guages, which have at most two SAE fea-tures.

Map 107.13: A cluster map combining nine fea-tures

Such cluster maps are thus a fairly direct rep-resentation of degrees of membership in a lin-guistic area. But of course, the cluster mapdirectly reflects the choice of features that arecombined, and this choice is always some-what arbitrary. Of the twelve features in § 2,only nine were selected here because informa-tion on the other three was incomplete. Ide-ally, the features of § 3 should have beenadded, too. But it seems to me that the mainresults of Map 107.13 would not be changed(this map can also be compared to the verysimilar map in van der Auwera (1998a: 823),which combines five adverbial features orfeature clusters). The most striking featuresof Map 107.13 are:

(a) The nucleus of Standard Average Euro-pean is formed by French and German (afinding that led van der Auwera (1998a: 824)to propose the term Charlemagne Sprachbundfor the nuclear area of SAE). In view of the

historical role played by speakers of thesetwo languages both in the early medieval his-tory of continental Europe and in the veryrecent attempt at European unification, thisis of course an extremely intriguing result.(b) The southern European languages (bothRomance and Balkan languages) are at leastas close, if not closer to the nucleus than thenorthern languages and English. This meansthat it is misleading to call SAE features“Western European features”, as is some-times done. It is true that the Slavic lan-guages in the east lack many SAE features,but the Balkan languages are generally moreSAE than Slavic, although they are not west-ern European.(c) England stands somewhat apart from theEuropean nucleus (as noted also by van derAuwera 1998a: 823), although it is closelyrelated genealogically to German and hasbeen thoroughly influenced by French. SinceEnglish is currently the dominant languagethroughout the world, it is worth pointingout its somewhat marginal status among itsEuropean sister languages.

It is important to keep in mind that the fea-tures on which Map 107.13 is based have notbeen selected randomly and are thus by nomeans representative of the morphosyntacticfeatures of European languages. They wereincluded precisely because they were knownto show a distribution that supports the SAEhypothesis. Thus, no claim is made that all(or even the majority of) features will show asimilar distribution. It is perfectly possiblethat we will some day discover anotherSprachbund, based on a different set of fea-tures, that has Russian at its core and extendsall the way to western Siberia in the east andcentral Asia in the south, but within Europecomprises only the Slavic, Balkan, and Scan-dinavian languages. This area would overlapwith SAE, but it would not contradict it.Thus, a language may in principle belong todifferent linguistic areas, and different lin-guistic areas may coexist “on top of” eachother. Since areal typology is only in its in-fancy, we do not know how common suchsituations are, but nothing in the logic of aSprachbund implies that the world should beexhaustively divisible into non-overlappingSprachbünde.

In fact, a number of smaller linguisticareas within Europe have been proposed inthe literature (apart from the Balkan area,whose importance is not doubted by anyone),




e. g. by Lewy (1942), Wagner (1959), Decsy(1973), Haarmann (1976), and Ureland (1985)(cf. also Wintschalek 1993 on a Volga-Kamaarea). Currently the most thoroughly studiedareas are the Circum-Baltic area (cf. Stolz1991, Dahl & Koptjevskaja-Tamm (eds.)2001) and the Mediterranean area (cf. Cris-tofaro & Putzu (eds.) 2000). However, nostrong claims about a Circum-Baltic or a Me-diterranean linguistic area seem to have beenmade as a result of these studies.

5. How did SAE come into being?

Linguistic areas arise through language con-tact, but precisely which contact situationgave rise to Standard Average European isnot immediately clear. And what is the sourceof the various Europeanisms: Who borrowedfrom whom? A full discussion of the socio-historical, cultural and sociolinguistic issuesis beyond the scope of this article, so I willrestrict myself here to mentioning just fivepossibilities:

(i) retention of Proto-Indo-European struc-tures and assimilation of some non-Indo-European languages to Indo-Euro-pean language structure;

(ii) influence from a common substratumof a pre-Indo-European population inEurope;

(iii) contacts during the great trans-formations at the transition from lateantiquity to the early Middle Ages inEurope;

(iv) the official language (Latin) and thecommon European culture of the Mid-dle Ages;

(v) the common European culture of mod-ern times, from the Renaissance to theEnlightenment.

The fifth possibility must be rejected becausea time depth of 300�500 years is not suffi-cient to account for grammatical common-alities of the kind discussed above. If lexicalsimilarities between the European languagesare discussed � for instance neoclassicalcompounding (socio-/paleo-/ortho-/demo-,-graphy/-logy/-cracy, etc.) or idiomatic struc-ture (e. g. ivory tower/torre d’avorio/Elfenbein-turm, as poor as a church mouse/pauvre commeun rat d’eglise/arm wie eine Kirchenmaus) �then the last several centuries are the appro-priate time frame for explaining the historicallinks, but the basic syntactic structures com-mon to SAE languages must be older.

The first possibility must be rejected be-cause the great majority of Europeanismsare innovations with respect to Proto-Indo-European. For instance, as far as we know,Proto-Indo-European did not have articles, a‘have’-perfect, “A and-B” conjunction, strictsubject agreement, particle comparatives, orrelative clauses with relative pronouns (cf.Lehmann 1974, Haspelmath 1998). With re-spect to Proto-Indo-European, and also withrespect to the oldest Indo-European lan-guages attested in Europe (Ancient Greek,Old Latin, Gothic), Standard Average Euro-pean is clearly an innovation.

The second possibility, a pre-Indo-Euro-pean substratum in Europe causing the SAEfeatures, would be extremely difficult to de-monstrate, but it might be worth pursuing. Itis intriguing to note that the geographicalspace occupied by SAE languages coincidesfairly precisely with the area of the Old Euro-pean hydronymy, i. e. the homogeneous layerof river names discovered by Hans Krahe(see Vennemann 1994 for recent discussion).Vennemann (1994) proposes that these OldEuropean hydronyms were not coined by anearly prehistoric Indo-European population,but by a pre-Indo-European people which hecalls Vasconic (the only surviving Vasconiclanguage being Basque). Furthermore, theOld European hydronymy is hardly attestedin the British Isles, where the Celtic lan-guages are spoken, i. e. they could not havebeen influenced by the Vasconic substratum.This is in perfect harmony with the well-mo-tivated hypothesis that the Celtic languagesacquired some of their striking features froma different substratum related to the Afro-Asiatic languages (Pokorny 1927�30, Gens-ler 1993).

The main argument against the substratumview is that the SAE features seem to be gain-ing ground too late for a pre-Indo-Europeansubstratum to have caused them. Some SAEfeatures appear only in the first milleniumCE, but also the earlier features usually comefairly late, so that the earliest records of Indo-European-languages in Europe still showtraces of the Proto-Indo-European patterns(e. g. causatives, relative clauses, locative com-parative, “A B-and” conjunction). If theseSAE features were caused by a substratum,then we should have much more evidence ofthe population speaking this substratum lan-guage. Moreover, a Vasconic substratum canhardly account for the SAE features becausemodern Basque is in most relevant ways verymuch unlike the SAE languages.




Of the remaining two possibilities, we canprobably exclude option (iv) (the influence ofLatin in the Middle Ages), because most SAEfeatures were absent in Latin and developedonly in the Romance languages. There areonly two features for which Latin influenceis a likely factor: negation and relative pro-nouns. In the case of these two features, thestandard languages sometimes show devia-tions from the vernacular dialects, so at leastthe written standard languages may havebeen influenced by Latin, the European writ-ten language par excellence for many centu-ries. Thus, non-standard English has con-structions like I won’t do nothing (‘I won’t doanything’), and similarly in non-standardGerman and French (cf. Haspelmath 1997:205). Analogously, Latin-type relative pro-nouns occur widely in the standard languagesof Europe, but vernacular speech often pre-fers relative particles (Lehmann 1984: 88,109). However, Latin probably only helpedto reinforce these structures in those lan-guages where they existed already indepen-dently as variants.

Thus, we are left with option (iii), the timeof the great migrations at the transition be-tween antiquity and the Middle Ages. Thisseems to be the appropriate time frame atleast for articles, the ‘have’-perfect, the par-ticipial passive, anticausatives, negative in-definites, nominative experiencers and verbfronting. The rise of these constructions canbe observed only with difficulty because theywere by and large absent in the written classi-cal languages but seem to be well in placeonce the vernacular languages appear in thewritten record toward the end of the firstmillennium CE (cf. also Fehling 1980). Thishypothesis derives some further plausibilityfrom the fact that language contact musthave been particularly intensive and effectiveduring the great migrations, and in the caseof French and northern Italian we have am-ple records of the lexical effects of these con-tacts. However, it is not so easy to fit featuressuch as particle comparatives, „A and-B”conjunction and relative pronouns into thispicture, because these features seem to havedeveloped around the middle of the first mil-lenium BC or even earlier (cf. Haspelmath1998). Of course, we must always reckonwith the possibility (or even likelihood) thatdifferent SAE features are due to differenthistorical circumstances, and the correct pic-ture is likely to be much more complicatedthan we can imagine at the moment, let alonediscuss in this article.

6. Abbreviations of language names

Alb AlbanianArm ArmenianBlg BulgarianBrt BretonBsq BasqueCz CzechDut DutchEng EnglishEst EstonianFin FinnishFr FrenchGae Scots GaelicGrg GeorgianGrk GreekGrm GermanHng HungarianIce IcelandicIr IrishIt ItalianKom KomiLat LatinLaz LazLit LithunianLtv LatvianLzg LezgianMar MariMlt MalteseMrd MordvinNnts NenetsNor NorwegianPol PolishPrt PortugueseRom RomanianRus RussianSAE Standard Average EuropeanSam SaamiSCr Serbian/CroatianSln SloveneSpn SpanishSrd SardinianSwd SwedishTat TatarTrk TurkishUby UbykhUdm UdmurtUkr UkrainianWel Welsh

7. References

Abraham, Werner. 1999. “Preterite decay as aEuropean areal phenomenon”. Folia Linguistica33.1: 11�18.Ameka, Felix. 1996. “Body parts in Ewe gram-mar”. In: Chappell, Hilary & McGregor, William(eds.). The grammar of inalienability: A typological




perspective on body part terms and the part-wholerelation. Berlin: Mouton de Gruyter, 783�840.

Bechert, Johannes & Bernini, Giuliano & Buridant,Claude (eds.). 1990. Toward a typology of Europeanlanguages. (Empirical Approaches to Language Ty-pology, 8.) Berlin: Mouton de Gruyter.

Beckman, Natanael. 1934. Västeuropeisk syntax:Nagra nybildningar i nordiska och andra västeuro-peiska sprak. [West European syntax: Some inno-vative constructions in the Nordic and other WestEuropean languages.] Göteborg: Göteborgs hög-skolas arsskrift.

Benveniste, Emile. 1960. “ ‘Etre’ et ‘avoir’ dansleurs fonctions linguistiques”. Bulletin de la Societede linguistique de Paris 55: 113�134. (Reprinted in:Benveniste, Emile. 1966. Problemes de linguistiquegenerale. Paris: Gallimard, 187�207.)

Bernini, Giuliano & Ramat, Paolo. 1996. Negativesentences in the languages of Europe: A typologicalapproach. (Empirical Approaches to Language Ty-pology, 16.) Berlin: Mouton de Gruyter.

Bossong, Georg. 1998. “Le marquage de l’expe-rient dans les langues d’Europe”. In: Feuillet (ed.),259�294.

Comrie, Bernard. 1998. “Rethinking the typologyof relative clauses”. Language Design 1: 59�86.

Cristofaro, Sonia & Putzu, Ignazio (eds.) 2000.Languages in the Mediterranean area: typology andconvergence. Milan: Franco Angeli.

Dahl, Östen. 1990. “Standard Average European asan exotic language”. In: Bechert et al. (eds.), 3�8.

Dahl, Östen. 1995. “Areal tendencies in tense-aspect systems.” In: Bertinetto, Pier Marco & Bi-anchi, Valentina & Dahl, Östen & Squartini, Mario(eds.) Temporal reference, aspect and actionality,vol. 2: Typological perspectives. Turin: Rosen-berg & Sellier, 11�28.

Dahl, Östen. 1996. “Das Tempussystem des Deut-schen im typologischen Vergleich.” In: Lang,Ewald & Zifonun, Gisela (eds.). Deutsch � typo-logisch, Berlin: de Gruyter, 359�368.

Dahl, Östen (ed.). 2000. Tense and aspect in the lan-guages of Europe. (Empirical Approaches to Lan-guage Typology-EUROTYP, 20�6.) Berlin: Mou-ton de Gruyter.

Dahl, Östen & Koptjevskaja-Tamm, Maria (eds.)(2001) Circum-Baltic languages. Vol 1�2. Amster-dam: Benjamins.

Decsy, Gyula. 1973. Die linguistische StrukturEuropas. Wiesbaden: Harrassowitz.

Dryer, Matthew. 1989a. “Large linguistic areas andlanguage sampling”. Studies in Language 13:257�92.

Dryer, Matthew. 1989b. “Article-noun order”. Chi-cago Linguistic Society 25: 83�97.

Dryer, Matthew. 1998. “Aspects of word order inthe languages of Europe”. In: Siewierska, Anna(ed.) Constituent order in the languages of Europe.Berlin: Mouton de Gruyter, 283�319.

Fehling, Detlev. 1980. “The origins of Europeansyntax”. Folia Linguistica Historica, 1: 353�387.

Feuillet, Jack (ed.). 1998. Actance et valence dansles langues d’Europe. (Empirical Approaches toLanguage Typology-EUROTYP, 20�2.) Berlin:Mouton de Gruyter.

Garvin, Paul. 1949. “Standard Average Europeanand Czech”. Studia Linguistica 3: 65�85.

Gensler, Orin D. 1993. A typological evaluation ofCeltic/Hamito-Semitic syntactic parallels. Ph. D.dissertation, University of California at Berkeley.

Gilligan, Gary M. 1987. A cross-linguistic approachto the pro-drop parameter. Ph. D. dissertation,USC.

Haarmann, Harald. 1976. Aspekte der Arealtypolo-gie: Die Problematik der europäischen Sprachbünde.Tübingen: Narr.

Haspelmath, Martin. 1990. “The grammaticizationof passive morphology”. Studies in Language, 14.1:25�71.

Haspelmath, Martin. 1993. “More on the typol-ogy of inchoative/causative verb alternations”. In:Comrie, Bernard & Polinsky, Maria (eds.). Caus-atives and transitivy. Amsterdam: Benjamins, 87�120.

Haspelmath, Martin. 1997. Indefinite pronouns.Oxford: Oxford University Press.

Haspelmath, Martin. 1998. “How young is Stan-dard Average European?” Language Sciences 20:271�87.

Haspelmath, Martin. 1999. “External possession inan European areal perspective”. In: Payne, DorisL. & Barshi, Immanuel (eds.) External possession.(Typological Studies in Language, 39.) Amster-dam: Benjamins, 109�135.

Haspelmath, Martin. 2001. “Non-canonical mark-ing of core arguments in European languages”. In:Aikhenvald, Alexandra & Dixon, R. M. W. (eds.)Non-canonical subjects and objects. (TypologicalStudies in Language) Amsterdam: Benjamins.

Haspelmath, Martin. To appear. “Coordination”.In: Shopen, Timothy (ed.) Language typology andlinguistic description. 2nd ed. Cambridge: Cam-bridge University Press.

Haspelmath, Martin & Buchholz, Oda. 1998.“Equative and similative constructions in the lan-guages of Europe”. In: van der Auwera (ed.),277�334.

Heine, Bernd. 1994. “Areal influence on grammati-calization”. In: Pütz, Martin (ed.) Language con-tact and language conflict. Amsterdam: Benjamins,55�68.

Heine, Bernd. 1997. Possession: Cognitive sources,forces, and grammaticalization. Cambridge: Cam-bridge University Press.

Jakobson, Roman. 1931. “K xarakteristike evrazij-skogo jazykovogo sojuza”. [Characterizing theEurasian linguistic area.] Reprinted in: Jakobson,




Roman. 1971. Selected writings, vol. I. The Hague:Mouton, 144�201.

Kahrel, Peter. 1996. Aspects of negation. Ph. D. dis-sertation: University of Amsterdam.

König, Ekkehard & Haspelmath, Martin. 1998.“Les constructions a possesseur externe dans leslangues d’Europe”. In: Feuillet (ed.), 525�606.

König, Ekkehard & Haspelmath, Martin. 1999.“Der europäische Sprachbund”. In: Reiter, Norbert(ed.). Eurolinguistik. Wiesbaden: Harrassowitz,111�127.

König, Ekkehard & Siemund, Peter. 1999. “Inten-sifiers and reflexives: a typological perspective.” In:Frajzyngier, Zygmunt & Curl, Traci S. (eds.) Re-flexives: Forms and functions (Typological studiesin language, 40.). Amsterdam: Benjamins, 41�74.

Kortmann, Bernd. 1997. Adverbial subordination: Atypology and history of adverbial subordinatorsbased on European languages. Berlin: Mouton deGruyter.

Lazard, Gilbert. 1990. “Caracteristique actanciellesde l’‘europeen moyen type’.” In: Bechert et al.(eds.), 241�53.

Lazard, Gilbert. 1998. „Definition des actants dansles langues europeennes”. In: Feuillet (ed.), 11�146.

Lehmann, Christian. 1984. Der Relativsatz. Tübin-gen: Narr.

Lehmann, Winfred P. 1974. Proto-Indo-Europeansyntax. Austin: University of Texas Press.

Lewy, Ernst. 1942. Der Bau der europäischenSprachen. (Proceedings of the Royal Irish Acad-emy, Vol. 48). Dublin: Hodges & Figgis.

Masica, Colin P. 1976. Defining a linguistic area.Chicago: University of Chicago Press.

Nedjalkov, Igor’ V. 1998. “Converbs in the lan-guages of Europe”. In: van der Auwera (ed.),421�455.

Nichols, Johanna. 1992. Linguistic diversity inspace and time. Chicago: The University of Chi-cago Press.

Nichols, Johanna & Peterson, David A. & Barnes,Jonathan (to appear). “Causativizing and decau-sativizing languages”. Ms., University of Cali-fornia, Berkeley,

Nikolaeva, Irina. 1999. Ostyak. Munich: LincomEuropa.

Nocentini, Alberto. 1996. “Tipologia e genesi del-l’articolo nelle lingue europee”. Archivio Glottolo-gico Italiano 81.1: 3�44.

Pisani, Vittore. 1969. “Tipi d’accento nelle linguedell’Europa”. In: Actes du Xe Congres Internationaldes Linguistes, Bucuresti, vol. II, 57�60. Bucuresti.

Pokorny, Julius. 1927�30. “Das nicht-indogerma-nische Substrat im Irischen”. Zeitschrift für celti-sche Philologie 16: 95�144, 231�266, 363�394; 17:373�388; 18: 233�248.

Ramat, Paolo & Bernini, Giuliano. 1990. “Area in-fluence versus typological drift in Western Europe:the case of negation”. In: Bechert et al. (eds.),25�46.

Siewierska, Anna. 1999. “From anaphoric pronounto grammatical agreement marker: Why objectsdon’t make it”. Folia Linguistica 33.1�2: 225�51.

Stassen, Leon. 1985. Comparison and universalgrammar. Oxford: Blackwell.

Stassen, Leon. 2000. “and-languages and with-languages”. Linguistic Typology 4.1: 1�54.

Stolz, Thomas. 1991. Sprachbund im Baltikum?Estnisch und Lettisch im Zentrum einer sprachlichenKonvergenzlandschaft. Bochum: Brockmeyer.

Stolz, Thomas. 1996. “Some instruments are reallygood companions � some are not: On syncretismand the typology of instrumentals and comita-tives”. Theoretical Linguistics 23.1�2: 113�200.

Stolz, Thomas. 2001a. “No Sprachbund beyondthis line! On the age-old discussion on how to de-fine a linguistic area”. Sprachtypologie und Univer-salienforschung 54.

Stolz, Thomas. 2001b. “Ordinalia � LinguistischesNeuland: Ein Typologenblick auf die Beziehungzwischen Kardinalia und Ordinalia und die Sonder-stellung von eins und erster”. In: Igla, Birgit &Stolz, Thomas (eds.) Was ich noch sagen wollte …Festschrift für Norbert Boretzky zu seinem 65. Ge-burtstag. Berlin: Akademie-Verlag.

Ternes, Elmar. 1998. “Lauttypologie der SprachenEuropas”. In: Boeder, Winfried & Schroeder,Christoph & Wagner, Karl Heinz & Wildgen, Wolf-gang (eds.) Sprache in Raum und Zeit: In memo-riam Johannes Bechert. Band 2. Tübingen: Narr,139�152.

Thieroff, Rolf. 2000. “On the areal distribution oftense-aspect categories in Europe”. In: Dahl (ed.),265�305.

Ultan, Russell. 1978. “Some general characteristicsof interrogative systems”. In: Greenberg, JosephH. (ed.) Universals of human language, vol. 4. Stan-ford: Stanford University Press, 211�248.

Ureland, P. Sture. 1985. “Sprachkontakt und Glot-togenese in Europa”. In: P. Sture Ureland (ed.)Entstehung von Sprachen und Völkern: Glotto- undethnogenetische Aspekte europäischer Sprachen. Tü-bingen: Niemeyer, 7�43.

van der Auwera, Johan. 1998a. “Conclusion”. In:van der Auwera (ed.), 813�36.

van der Auwera, Johan. 1998b. “Phasal adverbialsin the languages of Europe”. In: van der Auwera(ed.), 25�145.

van der Auwera, Johan (ed.). 1998. Adverbial con-structions in the languages of Europe. (EmpiricalApproaches to Language Typology-EUROTYP,20�3.) Berlin: Mouton de Gruyter.

van der Hulst, Harry & Hendriks, Bernadet & vande Weijer, Jeroen. 1999. “A survey of word pro-




sodic systems of European languages”. In: van derHulst, Harry (ed.) Word prosodic systems in the lan-guages of Europe. (Empirical Approaches to Lan-guage Typology-EUROTYP, 20�4.) Berlin: Mou-ton de Gruyter, 425�475.

Vennemann, Theo. 1994. “Linguistic reconstructionin the context of European prehistory”. Transac-tions of the Philological Society 92: 213�282.

Wagner, Heinrich. 1959. Das Verbum in den Spra-chen der britischen Inseln: ein Beitrag zur geographi-schen Typologie des Verbums. Tübingen: Niemeyer.

Whorf, Benjamin Lee. 1941. “The relation of habit-ual thought and behavior to language”. In: Spier,Leslie (ed.) Language, culture, and personality: Es-

108. Aire linguistique balkanique

1. Generalites2. Phonologie3. Systeme verbal4. Systeme nominal5. Autres unites6. Relations phrastiques7. Subordination8. References

1. Generalites

La linguistique balkanique est une disciplinerelativement recente, bien que la decouvertede traits communs entre les langues balkani-ques remonte a la premiere moitie du XIXe

siecle. Les specialistes (Asenova 1979: 5�45;Schaller 1975: 37�45) s’accordent a diviserl’histoire de la discipline en trois periodes:une periode preliminaire, ou l’on cherche aexpliquer les traits communs par l’influencedu substrat, une periode classique ou la lin-guistique balkanique acquiert ses lettres denoblesse grace a la publication en 1930 deLinguistique balkanique. Problemes et resul-tats de Sandfeld, qui represente la premieresynthese complete, et une periode moderne,marquee par le polycentrisme et l’internatio-nalisation des recherches (nombreuses revuesspecifiques et organisation de congres).

La linguistique balkanique ne consiste pasa juxtaposer des descriptions de langues di-verses dont le seul lien serait la contiguıtegeographique: il faut que ces langues for-ment une « union linguistique » (Sprachbund).Meme si certaines voix s’elevent encore pournier la realite de l’union balkanique (Andrio-tis & Kourmoulis 1968), la plupart des lin-guistes sont convaincus de son existence. En

says in memory of Edward Sapir. Menasha, Wis.:Sapir Memorial Publication Fund, 75�93. [Re-printed in Whorf (1956), 134�159.]

Whorf, Benjamin Lee. 1956. Language, thought,and reality: Selected writings of Benjamin LeeWhorf. Edited by John B. Carroll, Cambridge/MA:MIT Press.

Wintschalek, Walter. 1993. Die Areallinguistik amBeispiel syntaktischer Übereinstimmungen im Wolga-Kama-Areal. (Studia Uralica, 7.) Wiesbaden: Har-rassowitz.

Martin Haspelmath, MPI Leipzig(Deutschland)

effet, les traits communs sont trop nombreuxpour qu’ils soient le fruit du hasard. Il estvrai que les specialistes discutent encore dela notion de « balkanisme », que l’on definiraici comme un trait typologique propre a aumoins trois langues de l’union. Ce trait n’apas besoin d’etre unique en son genre (ainsi,l’article defini postpose existe dans les lan-gues scandinaves, le « redoublement » de l’ob-jet se retrouve dans les langues romanes); ildoit etre le resultat d’une convergence quiaboutit a un resultat identique ou quasi iden-tique, alors qu’il n’existait pas a des stadesplus anciens.

Les taches de la linguistique balkaniquesont consignees dans l’histoire de la disci-pline. Elles ont un triple aspect: synchronique(description) panchronique (extension) etdiachronique (formation et evolution). Bienque l’essentiel du travail descriptif sembleavoir ete acompli (la monographie de Sand-feld a ete completee, souvent amelioree, pardes centaines d’articles et d’etudes de detailqui ont permis d’accroıtre et d’approfondirles donnees), il reste toujours beaucoup afaire. L’etude de l’extension des balkanismesnecessite le recours a la geographie linguisti-que (ou linguistique areale) pour determineravec exactitude le lieu d’apparition de chaquebalkanisme et son extension reelle sur le ter-rain. Enfin, la perspective diachronique n’estjamais perdue de vue par les balkanologues,malgre les nombreuses difficultes auxquellesils sont confrontes, faute de documents ecrits.Trois aspects sont a prendre en considera-tion: 1) La genese de l’union linguistiquebalkanique; 2) La genese des balkanismes; 3)



The European linguistic area: Standard Average European (2001)

Documents