Are all agglutinative languages related to one another?
Prof. Dr. Alfréd Tóth
In linguistics, languages can be compared to one another either by genetic or by typological classifications. Genetic relationship means that all the languages compared are (supposed to be) genetically related to one another like the members of a family. An example is the Germanic language family, which contains amongst other languages German, Dutch, English, Danish, Swedish, the two Norwegians, Icelandic, Färöic etc. Typological relationship means that certain languages – that are not or not necessarily genetically related to one another – share certain (mostly syntactic) features. Examples are Biblical Latin, Korean, Chinese and Vietnamese because they are all topic-prominent (Tóth 1992).
Genetic classification
of languages goes back to Wilhelm von Humboldt (1767-1835) and his successors who became founders of comparative historical linguistics of the
Indo-European languages. But already von Humboldt, August Schleicher (1821-1868) and others introduced early typological classifications of languages
and suggested that typologically similar languages may also be genetically related. This was the basic reason why already very early Indo-European and
Semitic were compared to one another – because they are the only two big flectional (flexive) language families, and it was thought that this could
not be by chance. Nowadays, one differentiates at least 4 (mostly overlapping
and partially contradicting) sorts of typological
classifications:
- Morphological: analytic (ex.: English) – isolating (ex.: Chinese) – synthetic (ex.: most Indo-European languages) – fusional (ex.: Indo-European, Semitic) – agglutinative (ex.: Uralic, Altaic) – polysynthetic (ex.: Eskimo, Ainu) – oligosynthetic (ex.: Nahuatl)
- Morphosyntactic: nominative-accusative languages (ex.: Indo-European, Semitic) – absolutive-ergative languages (ex.: Basque, Eskimo-Aleut)
- Syntactic: according to word order (Subject-Verb-Object, i.e. SVO and all possible combinations)
- Pragmatic: subject-predicate languages (Indo-European, Semitic) – topic-comment languages (Chinese, Vietnamese) or both (Korean, Hungarian)
Because these categories are both overlapping and partially contradictory, some classifications turn out very odd. So is, e.g., Tibetan agglutinative but has Ablaut (apophony) like fusional languages (ex.: Engl. sing – sang – sung), but unlike fusional languages, Tibetan is ergative-absolutive. Verb-initial word order structures (e.g., VSO) are typical for Semitic languages (yet except Akkadian and Rhaetic), but Old Irish (Indo-European) also follows this type (and clearly not under Semitic influence). Most ergative languages are agglutinative and thus do not belong to the Indo-European languages, but Hindi, Punjabi, Kurdish and Ossetic are ergative and nevertheless Indo-European, although they are also characterized by fusion. On the other side, the “hyperflexive” Caucasian languages are not fusional, but agglutinative, etc.
Common sense in comparative linguistics is
still that there is no need for typologically related languages to be also genetically related and vice versa. The basic reason is the assumption that
languages may change their typological structure: “Wir neigen also zur Annahme, dass der
indogermanische Sprachbau auf dem Wege der
Überwindung eines primitiven flektierenden Typus entstanden ist, ohne jedoch den höher entwickelten agglutinierenden Typus erreicht zu haben/We thus
tend to assume that the structure of the Indo-European languages originated in overcoming a primitive flective type, but without ever having reached
the more highly developed agglutinative type” (Trubetzkoy 1939, p. 89).
Trubetzkoy thus assumes (contradicting most other linguists) that flexive languages are not the crown of creation, but an initial or intermediate stadium on the way to agglutination. Already Brunner (1969: 4) assumed that Proto-Indo-European was agglutinative – an assumption that was recently shared by Lehmann (2002). And looking at Tibetan that seems to have conserved its Sumerian heritage at least what concerns its grammar best of all languages rooting in Sumerian, is agglutinative with Ablaut like Sumerian was. Therefore, Trubetzkoy may be right despite the fact that nobody followed him: Ablaut is the basic feature for flexional languages, but flexion alone can also be expressed by agglutination. So, Sumerian was kind of hyper-characterized (as Tibetan still is), thus leading on one side into language families that concentrated on Ablaut and thus became flexive, and on the other side to language families that concentrated on flexion and thus became agglutinative.
From this standpoint, isolating languages in which particles fulfil the functions of affixes seem to be even a step further – this assumption also being in contradiction with most linguists. And it is surely not by chance that the isolating Malayo-Polynesian language family could be proven related to the flexive Semitic and Indo-European families (Brunner 1982).
Generally, the suspect arises that languages in a very early stadium were hyper-characterized also what concerns the other typological features: Since in ergative languages the object in an intransitive clause is marked by the same case as the subject in a transitive clause, the syntactic role subject, the semantic role agens and the pragmatic role topic do not coincide as they do in accusative languages. Thus, the accusative languages seem to show a certain linguistic economy that is typical for more developed languages, but not for early ones. The same seems to be true for topic-prominent languages, since topics can fulfill any syntactic and semantic function, even settings can be topic (e.g., the beginning of the famous Lili Marleen song), while in subject-prominent languages the subject is mostly identical with the topic or has otherwise to be marked by special markers or syntactic structures.
It is therefore not plausible at all to assume that the older a language – the simpler its structure. From our theoretical considerations as well as from the oldest testified language – Sumerian – it follows clearly that very early languages had all the possibilities together that have become differentiated in later stages of these languages following strategies of linguistic economy and leading to the known typologically differentiated languages and language families. And this seems to be true for all 4 typological structures mentioned above.
Relatively recently, few linguists resumed the question if there may be certain chances of genetictypological relationships: Hakola compared Finnish, Japanese, Mongolian, Quechua and Tamil and came to the following conclusion: “Examination of accidental CVC and CV correspondences among languages representing 5 large families of agglutinative languages found that comparison pairs had much more singularity between basic 100-word vocabularies than would have been possible by mere chance, supporting the hypothesis that those 5 language families were mutually related” (1989, p. 394).
Anttila stated: “In this way typology is hierarchically superior to genetic linguistics” (1989, p. 318). Anttila’s statement is of
particular interest. It means that languages can change, but only in the frame of their typological constraints. It follows that typological
structures are inheritable. This insight has huge consequences, since up to Anttila, e.g., the agglutinative structure of Indo-European languages like
Tocharian, Ossetic or Kurdish was explained either via substrate or adstrate effect: Tocharian, e.g., was allegedly agglutinative only because its
neighboring languages (Uighur, Old Chinese) were agglutinative. The same was assumed for Kurdish (influence of agglutinative Turkish) and Ossetic (via
agglutinative Caucasian languages), but since agglutination is inheritable, nothing stops us from assuming that Ossetic may have inherited its
agglutinative structure from its also agglutinating ancestorlanguages Scythian, Sarmatian, Alanian, Sacian and Massagetian.
It can be shown that
the distribution of many typological features of languages is not random but geographically (relatively) restricted. E.g., ergative-absolutive
languages show up basically in the Caucasus, in North America, Mesoamerica, Australia: Basque, Berber, Dyirbal, Eskimo-Aleut, Kurdish, Mayan,
Mixe-Zoque, Samoan, Tagalog and many other Austronesian languages, Sumerian, Tibetan, Caucasian without Kartvelian. Since agglutination is
inheritable, we may thus ask if the agglutinative languages are also concentrated in certain regions of the world. Unfortunately, since there is no
complete list of agglutinative languages (but cf. Shibatani/Bynon 1999), the following overview may be
incomplete:
- Uralic (Collinder 1957)
- Altaic (Turkic, Mongolian, Manchu, Korean, Japanese) (Ramstedt 1966; Poppe 1960; Sohn 2001; Miller 1971 [with review Menges 1974])
- Eskimo-Aleut (Mithun 1999)
- Paleo-Siberian (Chukotko-Kamchatkan, Yukaghir, Yeniseian, Gilyak) (Comrie 1981)
- Ainu (Tamura 2000)
- Tibeto-Burman (van Driem 2001; also some Chinese languages like Wu; Old Chinese?Clauson?)
- Basque (Hualde/Ortiz de Urbino 2003)
- Caucasian Languages (Klimov 1980)
- Punjabi (Bhatia 1993)
- Ossetic (Thordarson 1989)
- Kurdish (Wurzel 1997)
- Cushitic Languages (Saeed 1993)
- Bantu Languages (Guthrie 1971)
- Dravidian (Kirshnamurti 2003)
- North American Indian Languages (von Sadovszky 1996; Kroeber 1999)
- Mesoamerican Indian Languages (Campbell 1997)
- South American Indian Languages (Derbyshire/Pullum 1986)
- Malaysian (Lynch/Ross/Crowley 2002)
The following ancient languages were also agglutinative:
- Pre-Indo-European (Lehmann 2002; Greenberg 2000)
- Proto-Indo-European (Brunner 1969)
- Etruscan (Pfiffig 1969)
- Tocharian (Krause/Thomas 1960)
- Sumerian (Thomsen 1984; Edzard 2003)
- Elamite (Khacikjan 1998)
- Hurrian (Wegner 2000)
- Urartian (Diakonoff 1971)
- Hattic (Girbal 1986)
- Kassite (Balkan 1954)
- Gutian (Hallo 1957)
- Lullubi (Speiser 1931)
From this brief list, we can conclude:
- All known Mesopotamian languages (excluded the later Semitic languages like Akkadian, Rhaetic, Amoritic, Ugaritic, etc.) were agglutinative.
- The geographical distribution of the agglutinative languages is more or less identical with the languages that have been suspected in the past to be related to Hungarian and thus have been researched in my “Etymological Dictionary of Hungarian” (Tóth 2007b): Roughly speaking, they extend from the Ice Sea to the Southern Seas leaving huge “gaps” only in certain parts of India (e.g., no member of the Mon-Khmer family is according to my knowledge agglutinative).
Therefore, agglutination is not only inheritable, but agglutinative languages seem to cover a more or less coherent territory with a huge extension both in space and in time. Although not all languages are sufficiently documented, it is possible to show the genetic relationship of typologically related languages with Bouda’s concept of “Brückensprachen” (“bridging languages”) (cf. Bouda 1963). These are languages that connect both genetically and typologically related languages that are geographically (nowadays) distant. The concept of bridging languages is the more useful because, as already stated, languages can change their typological structure during their evolution. E.g., Old Chinese was agglutinative (as, e.g., Wu Chinese still is), while it is now isolating. The same may be true for the Mon-Khmer languages (cf. Shorto/Sidwell/ Bauer 2006, p. 590ss.). The special problem with India is that many of the hundreds of languages are not even researched yet.
All the facts mentioned point strongly in favor of our suspect that all agglutinative languages are genetically related.
In order to prove this hypothesis, I will proceed in two steps. First, I have collected all serious etymological studies that concern the genetic interrelationships between the above listed languages and counted the number of established word equations. From that, there can be no doubt that the phonetic proof that all these languages are related, has already be done. Second, I will apply Fokos-Fuchs’ (1962) catalogue of 25 syntactic and morpho-syntactic features that he used to prove syntactically the genetic relationship of the Uralic and Altaic languages, in order to demonstrate that Sumerian, Hungarian and the Dravidian languages share all features of Fokos-Fuchs’ catalogue.
As representative of Dravidian I chose Kannada (Jensen 1969). The reason why I did not chose any other agglutinative language, is because most linguists believe since McAlpin (1981) that the Dravidian language originate in Elamite, although McAlpin brings only 81 word-equations while there are many hundreds of Sumerian-Dravidian cognates. Now, Elamite is besides Sumerian the only sufficiently documented ancient agglutinative Mesopotamian language. Thus, if we succeed in proving that the Dravidian languages are also syntactically related to Sumerian, we will be allowed to assume that all agglutinative language may be related to one another on the simple reason because they all go back to Sumerian.