Varja Cvetko-Orešnik CDU81'1 University of Ljubljana 
THE o-GRADE OF THE INDO-EUROPEAN ABLAUT IN THE 
LIGHT OF NATURALNESS 

Jf the (probably obsolescent) point of view is accepted that the o-grade ofthe Proto-Jndo-European ab/aut arosefrom the more basic e-grade under lack ofword accent, Naturalness Theory can stale the assumptions which make it possible to predic! the distribution ofthe two grades in dependence upon the accent. One assumption is, >nat (+accented, -accented) / vowel, the other assumption is, >nat (e-vowel, o-vowel). Using "markedness agreement", the consequence is asfollows. lfthere is any difference between the e-vowel and the o-vowel, such that one kind ofvowel is accented, and the other kind of vowel is unaccented, it is the e-vowel that tends to be accented, and it is the o-vowel that tends to be unaccented. 
This method can be used to check ifsome presupposed language state is really possible in natu­ra! languages. 
This paper discusses an aspect ofthe old (probably obsolescent) text-book wisdom conceming the Indo-European ablaut, namely the (once generally accepted) assump­tion that the o-grade is younger than than the e-grade: the o-grade arose from the e­grade whenever the latter was not accented (under certain additional conditions). The aim of this paper is to show that accented e-grade and unaccented o-grade constitute a possible state of affairs in natura! languages. To this purpose, 1 will make use of Naturalness Theory as currently practised especially in Austrian linguistics (see Dressler 2000), and in my own surroundings at the University ofLjubijana (cf. Orešnik 2001). Naturalness Theory operates with naturalness scales which ascribe greater or smaller naturalness to language units and categories. One of the leading theoreticians in the field ofnatural morphology, Austrian-based Willi Mayerthaler, distinguishes two kinds ofnaturalness, sem-and sym-naturalness (cf. Mayerthaler 1981 etc.). Since 1 will make use of sem-naturalness only in the present paper, and as the latter is very similar to traditional markedness, I will refrain from discussing the difference between sem-and sym-naturalness. 1 will use the bare term naturalness, and set up the follow­ing approximate equations: more natura! = less marked, less natura! = more marked. The predicate natura! will be abbreviated as 'nat' in scales. 
The essential difference between markedness and naturalness lies in the way they are determined. Markedness is determined ( computed) using the means at the dispos­al within the language whose units/categories are being dealt with. For instance, it can be assumed that unit A is less marked than unit B, ifunit A's sphere ofusage is greater that unit B 's sphere of usage. Other ways of ascertaining markedness are comfortably enumerated e.g. in Andersen 2001. Contrariwise, the determination of naturalness is not limited to the means at the disposal in a single language, and broader data, for example typological data, are made extensive use of. For instance, if category A is wider-spread cross-linguistically than category B, it can be assumed that category A is more natura! than category B. Thus naturalness allows more ways of computation than markedness. Therefore I will make use of naturalness in the present paper. 
The principled difference between markedness and naturalness is only apparently great. In both instances, the essential act is to determine indirectly how the human brain evaluates the relative difficulty, the relative stability etc. of any units and cate­gories of a language. 
For the purposes of the present paper I propose to suggest two new naturalness scales. One is, >nat (+accented, -accented) / vowel. The other scale is, >nat (e-vowel, o-vowel). I proceed to adduce argumentation in support of the two scales. 
The scale, >nat ( +accented, -accented) / vowel asserts that an accented vowel is more natura! than an unaccented vowel. The term accent is here used neutrally with respect to the kind ofaccent; the latter can be dynamic or tonemic. The scale can be sup­ported with typological facts. In some languages, the word is only or almost only mono­syllabic. Vietnamese is one of such Ianguages. It distinguishes accented and unaccent­ed words. The class of unaccented words is limited, the class of accented words is not limited. Obviously, accented words are basic in this language (and in fact in any lan­guage). It follows that the accented monosyllabic word is prototypical in natura! Ian­guages. Given that in languages whose words are monosyllabic any difference between an accented word and an accented vowel can be neglected, it can be asserted that the accented vowel is the prototypical vowel in the languages of the world. This jibes well with the known fact that the class of accented vowels of many Ianguages is richer than the class of unaccented vowels. It has been observed within the theory of markedness that what is less marked often displays richer forms than what is more marked. For instance, it is expected that, in a language having case forms, the singular of nominal parts of speech shows more cases than the non-singular of those parts of speech. 
The suggestion that monosyllabic accented words are prototypical can be support­ed in a negative way as well. There are no languages displaying only one structural type of accented words such that that structural type would not be monosyllabic. 
My other scale is, >nat (e-vowel, o-vowel), i.e. an e-vowel is more natura! than an o-vowel. This scale can also be supported with typological data. As emphasized by Roman Jakobson 60 years ago (Jakobson [1940] 1969, 75-6 with references), some languages have e-vowels and Jack o-vowels; on the other hand, there are no languages having o-vowels and lacking e-vowels. Jakobson also drew attention to the following: during language acquisition, the child acquires e-vowels earlier than o-vowels. 
l now propose to link the two scale with each other. I will use the principle of "markedness agreement" as formulated by Andersen 1972. (To be sure, Andersen refers to the work of the Slavic scholar František Mareš. However, it is a fact that "markedness agreement" became known intemationally because of Andersen and in his interpretation and formulation.) Andersen's principle has remained uncontested; Andersen 2001 repeats it and adds new exemplification. According to "markedness agreement", an instance of less marked tends to associate with another instance of less marked, and an instance of more marked tends to associate with another instance of more marked. Transferred to naturalness, this means: an instance of more natura! tends to associate with another instance of more natura!, and an instance of less natura! tends to associate with another instance of less natura!. 
The principle dictates that the accented vowel is linked with e-vowel, and that the unaccented vowel is linked with o-vowel. 
However, since the link between the accented vowel and the e-vowels and the link between the unaccented vowel and the o-vowels are not the only possible links within the two pairs, other links being conceivable (for instance, between the vowel colour and the phonetic environment of the vowel), the preceding paragraph entitles me to suggest only the following two cautious alternative formulations (a) and (b) (whose format has been taken from Janez Orešnik's work, see Orešnik 2001 passim): 
(a) 
Ifthere is any difference between the accented and the unaccented vowel, such that one vowel is an e-vowel, and the other vowel is an o-vowel, then it is the accented vowel that tends to be an e-vowel, and it is the unaccented vowel that tends to be an o-vowel. 

(b) 
If there is any difference between an e-and an o-vowel, such that one of them is accented, and the other is unaccented, then it is the e-vowel that tends to be accent­ed, and it is the o-vowel that tends to be unaccented. 


I suggest that it can be seen from (a-b) that the link between the accented vowel and the e-vowel as well as the link between the unaccented vowel and the o-vowel are natura! in the sense that one would not be astonished if a language were found dis­playing exactly such a configuration of accentedness and vowel colour. In part, such a configuration can be observed (indirectly through reflexes or even directly) in some old Indo-European languages. 
It is known that comparative linguists reconstructing prehistorical states often address the question as to whether the reconstructed state is a possible state in natura! languages. Ifthey succeed in finding some natura! language in which the analogue of the reconstructed state is realized, the answer to the question is positive. However, it the linguist fails to find such a natura! language, the question arises as to how to sup­port any assertion that the reconstructed state is a possible state in natura! languages. I suggest that the linguist could, at least in some cases, profit from the method described above. 
To avoid misunderstandings, let me repeat that the above method helps to describe a reconstructed prehistorical state as a POSSIBLE state in natura! languages. The method does NOT assert that some given reconstructed prehistorical state is true/cor­rect. Consequently my paper should not be construed as advocating certain aspects of some (obsolescent?) theory conceming the Indo-European ablaut. 
References 
ANDERSEN. Henning. 1972. "Diphthongization". Language 48, 11-50. 
ANDERSEN. Henning, 2001. "Markedness and the theory of linguistic change". In press. 
Boou. Geert, Christian LEHMANN and Joachim MUGDAN (eds.), 2000. Morphologie: ein internationales 

Handbuch zur Flexion und Wortbildung. Volume l. Berlin, de Gruyter. DRESSLER, Wolfgang U., 2000. "Naturalness". In: Booij et al. eds, 288-96. JAKOBSON, Roman, [ 1940] 1969. Kindersprache, Aphasie und a/lgemeine Lautgesetze. Frankfurt/M, Suhrkamp. MAYERTl!ALER, Willi, 1981. Morphologische Natiirlichkeit. Wiesbaden, Athenaion. OREŠNIK, Janez, 2001. A predictable aspect oj(morpho)syntactic variants. Ljubljana, Slovenska akademija 
znanosti in umetnosti. 
Povzetek 
OJEVSKA STOPNJA PRAINDOEVROPSKEGA PREVOJA 
V LUČI JEZIKOVNE NARAVNOSTI 
Če se postavimo na nekdaj splošno sprejeto (a danes verjetno zastarelo) stališče, daje tisti prain­doevropski ojevski vokalizem, ki je v okviru preglasa nastal iz osnovnejšega ejevskega vokalizma, rezultat naglasnih razmer v besedi, namreč daje o nastal iz e-ja, kadar slednji ni bil naglašen, se dajo v okviru teorije o jezikovni naravnosti izreči predpostavke, na podlagi katerih se napoveduje raz­vrstitev ejevskega in ajevskega vokalizma glede na naglas. Ena predpostavka je >nat (+naglašen, -naglašen)/samoglasnik. Druga predpostavka je >nat (ejevski, ojevski)/vokalizem. Zdaj po "ujema­nju zaznamovanosti" sledi: če je kak razloček med ejevskim in ajevskim vokalizmom, tako daje ena vrsta vokalizma naglašena in druga vrsta vokalizma nenaglašena, potem teži k naglašenosti ejevski vokalizem in k nenaglašenosti teži ojevski vokalizem. 
Na ta način se da preverjati, ali so neke predpostavljene jezikovne razmere v naravnih jezikih tudi res možne.