Technical terms used in the description of a language can be broadly divided into three groups : (a) terms which truly belong to general linguistics, such as phoneme, morpheme construction, an idiom, lexicon, category, level; (b) terms which can be defined within general linguistics, if at all, only in a provisional manner, such as vowel, accent, word, verb, gender, attribute ,interrogative, emphatic, proper name, and (c) terms which really fall outside linguistics, but which are transferred from the practical world to things linguistic by a kind of metonymy, such as pitch, bilabial, before, after, future, negative, honorific, nursery (as in, nanny is a nursery version of nurse). The term case belongs to the second group.


When controversy surrounds the use of BASIC TERMS (our group a), it takes on a peculiar character.  A linguist will proceed to define a basic term according to the kind of theoretical model he adopts.  Indeed one of the principal features that differentiate any school of thought in descriptive linguistics is its choice of basic terms and the way they are defined.  But once we agree on them, we must rigorously apply them to any language that we may care to describe.  They are the only true "universals' in language.


            The VARIABLE TERMS (group b) should be approached in an altogether different manner.  They have to be defined afresh for each language - what is a vowel in one language need not be a vowel in one language need not be a vowel in the next language; and the same goes for accents, words, and the rest.  In turning its back on the mirage of a "universal grammar", modern linguistics enjoins us to be perfectly open-minded in describing any given language.  Now does this mean that we can never compare two languages and decide how similar they are as to structure-type? Does it mean that a typological as distinct from a historical comparison of language is an unattainable goal? Does it mean that a linguist can be perfectly arbitrary and irresponsible in what he decides to call vowel, word, gender, and so forth? It is to necessary to run to these extreme and rather despondent conclusions.  Given the theoretical model, a variable term will still have not one definition but a set of definitions - a series of definitions for these languages in describing which we have use for the term concerned rounded off by a tentative inductively arrived at general definition.  The general definition will form then on merely inform the linguist what are the sort of things for which the term is available, so that he does not have a hunt for a new term for each new language.  A clear distinction between basic terms and variable term will help us to avoid many a fruitless controversy on a terminological issue.


            Group (c) we may label as SUBSTANTIAL TERMS as against the first two groups, both of which are essentially formal in character.  The formal edifice that is language is anchored to our everyday "real" world at two ends-the phonetic and the semantic. The muscular wiggles, the waves in the air, the auditory sensations are all nonlinguistic events describable in terms that have no reference to the fact that these events describable in terms that have no reference to the fact that these events often happen to be related to formal entities i a language such as phonemes, morphemes, words, gender, etc. The same could be said about the thousand and one things in the practical world which may at short notice become the objects of attention for the speakers of a given language.   When a formal category is specifically and obviously related to some real-world fact, the linguist will find it convenient to label it after that semantic context. He may as well have used numbers of the letters of the alphabet-indeed he often does (witness, first person), but there is harm in choosing a less colourless term like masculine, so long as one does not expect all masculine nouns to refer to male organisms.  It is like providing a rough identificatory gloss, such as – ghor**ā   (Hindi) 'horse', although ghor**ā may also sometime stand for 'trigger of a gun'.  All this also applies, mutatis mutandis, to substantial terms at the expression and like pitch or before.


            Bearing in mind all the reservations implied in the foregoing discussion, we can proceed to give a tentative general definition of the variable term case, before turning to case in Marathi :


            If a language happens to have nouns (another variable term!) and if these nouns enter into regular paradigms (a basic term) such that the place occupied by a form in the paradigm marks the constructional relations the form enters into with other forms in the sentence in which it occurs, then the class of such mutually exclusive markers may be called case-markers.


            When we try to make use of this term in describing Marathi, three classes of markers offer themselves as suitable candidates to being called case.  Although in defining (Marathi) case, I have denied the name to two of them, the situation is worth examining as a whole.  Apparently closely analogous questions can be raised about other Indian languages and a clarification of this kind will be helpful in avoiding the pitfalls created by discrepant terminologies standing in the way of an intelligent comparison between these languages.


            Marathi throughout this paper means educated colloquial Marathi spoken in Poona in mid-twentieth century.  It will not be within our scope to bring in dialectological or historical material.  I have used myself as an informant, but I have extensively checked things with my fellow speakers.  All forms will be cited in phonemic transcription (enclosed within slant lines). Names of morphemes will be enclosed within brace brackets. No attempt is made here to present an exhaustive (and possibly exhausting) account of all minutiae and exceptions that have no bearing on our main concern.



            Let me now proceed to describe the relevant facts of Marathi grammar in such a manner as not to prejudge the issue of what may be suitably called case and of how then to replace the deliberately vague and tentative definition given earlier by a more rigorous one valid for Marathi.


            As a rule a Marathi stem belongs to one of three major classes depending on its total privileges of occurrence-especially in formations that are longer than a stem and shorter than or equal to a word.  These classes are nouns, verbs, and indeclinables.  Occasionally, however, a stem may belong to two or three classes by virtue of grammatical homonymy. (This last term is JESPERSEN'S and more apt than BLOOMFIELD'S class-cleavage).  Nouns are further divided into substantives and adjectives according to their privileges of occurrence in formations longer than a word.


            The regular paradigm in which a noun stem enters can be set forth most conveniently in the form of the following diagram (which reads from top to bottom):


A Noun Stem



With its appropriate Gender M, F, N



Either one of two Numbers – Sg., Pl.





Direct form                   Vocative form                           Oblique form             Class Y markers



                                                                                    Class X markers







Ready to enter into larger constructions


            Note that class Y markers (led to by a dotted arrow in the diagram) are not quite a part of the regular paradigm, since only a few noun stems take these and then too not all of them.






To take a concrete example, the stem /ghod))-/ 'horse' may be either M, F, or N.






Sg. Direct



ghod))ə̃ :

Sg. Vocative




Sg. Oblique




Pl. Direct




Pl. Oblique

ghod))ya (n)

ghod))ya (n)

ghod))ya (n)



            The plural oblique takes the /-n-/ only when either class X markers or the vocative particle /-o/ follows.  (Thus / ghod))yano ~ ghod))yanno/serves as the plural vocative).


            To the oblique forms—singular or plural-we can add any one out of a whole class of X markers, which come in all shapes and sizes.  Some can themselves be declined like noun stems, others not.  Some are single morphemes, others are complex forms exemplifying various constructions.  Some are phrase-bound and tend to have very little assignable content. Others are phrase-free and may more justifiably be treated as independent stems in their own right-the grammatical fusion with the noun in oblique is so incomplete that they tolerate the insertion of emphatic particles like /-c/ 'only', /-hi/ 'also' just after the noun in oblique.  A sample list follows :


            (1) Indeclinable – (la) Bound: /-ni/ 'agent, instrument' (after singular oblique/-ni ~ -nə̃:/); /-la/ 'object, recipient, destination' (after plural oblique /-la ~ -na/); /-s/ 'ditto' (more restricted in occurrence than the preceding); /-hun/ 'from, than'; /-ši/ 'near, with'; /-niši/ 'along with'; '/-pekša/ 'than'; /-gәt/ 'like, as'; -/pəryyənt/ 'up to, until'; /-kəd))e/ 'towards'.



                (1b) Free: /-ãt/ 'in' (regularly with loss of vowel after oblique); /-šivay/ 'without'; /-səhə/ 'accompanied (by)'; /-vər/ 'on, over, above'; /-зl$,,/ 'near'; /ulət,,/ 'opposed (to), against'; /-ad,,/ 'concealed (by), behind'.


            (2) Declined like noun stems--(2a) Bound : /-ca-či-cə̃ : / 'of, 's' (M, F, N respectively); /-зokta - зokti - зoktə̃:/ 'suited to'.


            (2b) Free: /-vegl,,a-vegl,,i-vegl,,ə̃:/ 'different (from) apart (from)'; /-ãtla/ etc. 'inner (to)' (with loss of vowel after oblique); /-sarkha/ etc. 'like, as'; /-purta/ 'enough (for)'.


            Coming to class Y markers we find that the stem / ghod))-/ happens to take none of them. A complete list follows roughly in order of list-frequency in relation to noun stems which take them. Note they are all bound and, with one exception, indeclinable :


            /-ĩ/ 'place, time', as in gǝl),,ĩ/ (/gǝl),,a/ 'throat, neck'), /ghǝrĩ/ '(at) home', /ratri/ (/ratrә/ 'night'), etc.


            Second/-ca –či -cǝ̄̄:/ 'quality', as in (з akca/(/ з aga/'place'), /ghərca/ 'coming from, belonging to home' (cf. /ghəraca/ with the first /-ca/ 'of the house'), /hacca/ (/hat/'hand'), etc.


            /-un ~ -nǝ͂:/ 'from', as in /ghərun ~ ghərnǝ͂:/,  /hatun~hatnǝ͂:/, etc.


            /- e͂͂/ 'instrument', as in /əndaзe͂͂/ (/ənda з/ 'estimate'), /bəl,,/  (bə!/ 'force'), etc.


            Second /-ĩ/ 'instrument', as in /paī/ 'on foot' (/pay/ 'foot'), /dhəd**pənī/ (dhəd**pənī/ 'undisturbed condition'), etc.


            /-il/ 'belonging to, from', as in /ghəril/ (colloquially /ghərãtla/ is more common), /daril/ 'not belonging to home' (colloquially /darca/ is more common; /dar/ 'door'), etc.


            /-ã/ 'place, time', as in /payã/ 'at the feet', /divsã/ 'by day' (/divəs/ 'day'), etc.


            The Y markers, it may be noted in passing, frequently participate in the formation of complex markers of class X, as in:


            /ãtun/ 'from inside', /vəril/ 'that which is on the top of', /pud**hẽ ~ pud**h ẽ : / 'in front (of)'.


            It may be urged that class Y markers can be dismissed as derivative endings.  Indeed they could be, but for one consideration.  Compare the following :


(1)   /candən,,i + rátrə/ 'moonlit night'.

(2)   /candən,,ya + rátri + la/ 'to a moonlit night'.  Note that both the noun stems are in oblique, while only the substantive takes the class X marker /-la/.

(3) /candən,,ya + rátri/ 'on a moonlit night'.  Note how the adjective takes an oblique in recognition of the Y marker /-i/ added tot he substantive.  If we were to consider /ratri/ to be merely an adverb derived from /ratrə/, this behaviour of the adjective will be without a parallel.  On the other hand, if we regard (2) and (3) as parallel formations having the same relation to (1), the adjectival oblique makes more sense.


        To summarize, in the noun paradigm three classes of endings satisfy the tentative definition of case given earlier :

(a)    The class constituted by {direct}, {oblique}, and {vocative};

(b)   Class X markers; and

(c)    Class Y markers.


A noun stem takes either (a) alone, or (a) followed by (b), or (c) alone.  It is clear that (a) and (b) cannot both be case-endings, since they occur together.  It is also clear that we have to decide for the classes as wholes, unless we can find some acceptable criterion to accord the status of case to some members of a class and withhold it from other members of the same class.  The procedure of traditional Marathi grammars whereby an arbitrary list cutting across our three classes is drawn up can be hardly justified.  The list runs as follows :


Traditional case


From (a)

From (b)

From (c)


I (nominative)






II (accusative)



-s, -la,









III (instrumental)



-ni, -ši,


- ẽ, -i2



IV (dative)



-s, -la,




V (ablative)






VI (genitive)



-ca, etc.

-ca, etc.


VII (locative)




-i1, -ã









            The two endings marked with an asterisk are now obsolete.  Note that the distinction between the two /-nẽ:/s (alternants to /-ni/ and to /-un/ respectively), and between the two /-ca/ series is ignored; that the endings {direct}, /-s/, and /-la/ are artificially split into a pair of homonyms each; and that the particle /-o/ (commonly cited as/-no/ is treated as a case-ending though it is an alternant of /ho/ (as in /mitrə +hó/ 'friends!', /cəlá + ho/ 'do come along!').  The several endings put under the same case are by no means freely substitutable for one another, nor are they in complementary distribution.  All this has no apparent motive except that is showing that Marathi, like a good daughter, has the same eight cases as Sanskrit has!


            Nothing has turned up in this discussion that will show that the tentative definition of case forces any particular choice on us.  We clearly need an additional criterion.  That is provided by another variable terms –INFLECTION.  A tentative general definition of inflection may be offered as follows:                                                                                                                                                                                    


            When the respective paradigms into which all the members of a (tentatively proposed) stem-class enter are matched to each other in respect of form and distribution, the following things (arranged roughly in order of their importance) may be observed :


(i)                  all or nearly all stems agree in being combinable with a set of markers;

(ii)                the distribution of the stem-plus-marker formation does not match that of some other simple stem (of the same class of different) so much as tie in with some syntactical relation;

(iii)               the presence of the marker closes the word;

(iv)              the set is small, compact, well-ordered, and not open-ended;

(v)                the markers are bound and determined by the stem.


When the paradigm or some specifiable subset of it fulfils most or all of these conditions, the paradigm or the subset concerned is the inflectional set for the stem-class in question (and may enter the definition of the proposed stem-class).


Let us apply these tests to the three classes of markers. Class Y disqualifies on the very first count.  Class X fails on the fourth count, and is surpassed by the other two on the third and the firth counts.  The set constituted by {direct}, {oblique}, and {vocative} comes nearest to fulfilling all the conditions. Case in Marathi is, therefore, the set of mutually contrasting and exclusive morphemes {direct}, {oblique}, and {vocative}. Classes X and Y we group together as POSTPOSITIONS –which is not a form-class so much as the second position in the axis-and-postposition construction, which can be occupied by any out of a large heterogeneous group.


In presenting the formal details of Marathi noun declension, gender, number, and case cannot be kept apart.  If we ignore the pronouns and other marginal cases, the principal declensional patterns are follows :


Masculine stems – Type 1: /ghod,,a/ 'horse, stallion', direct sg.; /ghod,,e/ direct pl.; / ghodya/oblique sg. and pl.


Type 2 : /vagh/ 'tiger', direct sg. and pl.; /vagha/ oblique sg. and pl.


Feminine stems – Type 1: /šal,,a/ 'school', direct sg. and direct and oblique pl.; /šal,,e/ oblique sg.


            Type 2 : /vat,,/ 'way, direct sg.; /vata,,/ direct and oblique pl.; /vat,,,e/ oblique sg.


            Type 3 : /rit/ 'method, customs', direct sg.; /riti/ direct and oblique pl. and   oblique sg.


Type 4 : /kal,,i/ 'black', direct sg.; / kal,,ya/ direct pl. and oblique sg. and pl.


Type 5 : /ghod,i/  ‘mare’, direct and oblique sg.; /ghod,ya/ direct and oblique pl.


Neuter stems – Type 1 : / з had,/ 'tree', direct sg.; / з had, ə̃ : / direct pl.; /з had,,a/ oblique sg. and pl.


Type 2 : /ghod,, ə̃̃:/ 'horse indifferently male or female; indifferent horse, nag', direct sg.; / ghod,,i/ direct pl.; / ghod,,ya/ oblique sg. and pl.


Notes : (1) The vocative singular is the same as the oblique singular except for feminine stems of types  4 and 5 (/kal,,e~ kal,,i/ and /ghod,,e/ respectively). (2) There is no vocative plural as such. (3) The oblique plural takes an additional /-n-/ when followed by a postposition (class X) or the vocative particle /-o/. (4) Feminine type 4 stems are all adjectival and type 5 all substantival. (5) Many adjectives come in triplets – M-1, F-4, and N-2; example: / kal,,a/, /kal,,i/, /kal,,ǝ̄:/ 'black'.




            Finally we may point out that in verbal paradigm case-morphemes show up in two places :


(1) Verbal stems regularly take the extensions /-t-/, /-l/, /-lel-/, /--/, /-ar-/ese extended stems not only take further part in the verb-inflection set but also function as noun-stems.

(2) Extended stems ending in /-t-/, /-l-/, /-av-/ have special oblique forms in which neither gender nor number morphemes enter; examples :


'/to + з áte+ vel,,i/ 'he going time-at, i.e. at the time of his going'.

'/to + gélya +vər/ 'he gone upon, i.e. upon his going'.

'/to + з áyči +vel,,/ 'he going-of time, i.e. the time of his going'.


The description is complete. But two possibly doubtful points may be cleared up before concluding.  The first concern the morphenic status of {oblique}.  Does it ever contrast with the other two cases? Does it ever occur without any postposition following? Could we not possibly regard oblique forms as grammatically conditioned allomorphs ? The contrast between / ghərca/ with the second /-c/ and / ghəraca/ ~ / ghərá + ca/ with the first /-c-/ has already been cited above while listing class Y markers.


The oblique form can occur, moreover, with no postposition following in the following cases :

(i)          the adjective in oblique when the following substantive is oblique or followed by class Y postposition :

/kal,,,ya + ghód,,i + vər/ 'on a black mare'.

/candən,,ya+ rátri/ 'on a moonlit night'.


(ii)        the first and the second noun taking separate obliques in certain pseudo-compounds :

/k ә ̓̓pd,,ya + ləttya + sat,,hi /'for clothes and things' (oblique plural; compare direct singular / k ә ̓̓pd,,a + lətta/).


(iii)       in a few steriotyped expressions, like :

/déva + ghəri/ 'God-obl. house-in, i.e. in heaven'.

/lóka + sange + . . ./ 'other-people-obl. tell...' (beginning of a proverb).

/paht,,,e/ 'at daybreak'.


In the face of this it will be difficult to show grammatical complementation; for, the larger the frame within which to define it, the more tenuous the complementation becomes – constantly rendered precarious by such "accidental" contrasts like :


/ti + candәn,,ya +rátri+cәmәkte/ 'it (fem.) shines on a moonlit night'.

            /ti + candәn,,ya +rátri+cәmәkte/ 'that star shines at night.'


            The second doubt concerns the way this analysis undermines the easy "isolability" that we associate with the word.  In most "word-using" languages, the word is the most important stopping point between the smallest unit in grammar—the morpheme- and the largest – the sentence.  But in Marathi that distinction apparently goes to the PHRASE. Once we bear in mind that word is a variable term, this need not be a disturbing conclusion.




            This was published in Indian Linguistics 20:131-9, 1959 (being Tumer Jubille Volume, Vol.2, Published October 1959)