ABSTRACT: a propos: Indira Y.Junghare.  Restrictive relative clauses in Marathi.  IL 34:4. 251-62 (1973).  Her analysis is good of its kind, but unsatisfactory in certain important ways.  an alternative analysis (section i) covering roughly the same territory is presented. Then follows (section ii) an account of relative clause abbreviation.  Possible extrapolations (section iii) of the analysis presented so far are sketched -to adjectival and adverbial relatives, multiple relatives, sentential nominalization, and, finally, to the so-called nonrestrictive relatives.  Certain methodological considerations (section iv) now suggest themselves.


            Indira Y.Junghare’s analysis of ‘restrictive relative clauses in Marathi’ (1973) is a good example of its kind – and i mean not only good as an example of its kind – and i mean not only good as an example but good as analysis.  But I certainly do want to suggest that it does not exemplify a very good kind of analysis – it is typical of a good deal of current transformational –generative work on Indian languages.  But I am anticipating.  It will be simpler to start with presenting an alternative analysis whole cloth – analysis which covers roughly the same territory.




The Starting Point


It is not necessary here to concern oneself as to how and at what point a Noun Phrase is thrown up in the course of specifying a Sentence.  Let it simply be assumed that it is so thrown up and has an S-node in it which triggers off the process of Relativization.  The structural description that need be assumed at the start, therefore, is as follows:

(Operand 1) Relativization Operand




X         NP       Y        




                              S1                              NP΄              



X1   NP1        Y1             Correl      NP΄   Residue  



                 Rel      NP1       Residue




It will be noticed that we have simply refrained from taking up any position on whether Marathi has a VSO or SOV or SVO order – that is, on whether X or Y or at least one of them is necessarily non-null.  But we cannot refrain from having some notion as to how NP is specified in Marathi.


(Grammatical Formation Rule 1) NP Specification


            NP →   Det N

                         S NP


Conditions: (1) If S is present the operands NP corresponds to the NP in Opd 1  above.

(2) Det may or may not yield a Stem; and the same goes for N.  But both Det and N cannot be Stem-less at the same time.


            GFR 1 is thus capable of yielding strings like Det N, S Det N, S S Det N, S S S Det N, and so on.


(Opd 1 continued)


            Conditions (1) Rel = Det Indef [/k-]

                                                     Def [t-/ 1st/2nd/Proper]

                              (2) Correl = Det [Def [t-] [(1st/2nd/Proper)]

                              (3) If an S-nәde is present in NP΄ Residue, it has already

been subjected to Relativization.

                              (4) If  NP is a Predicative Complement in S, then Re1 = Det

                                    [Indef  [] and NP΄ Residue = ]N [G No C [dir ] … ]


            Note that G, m, f, n; No, sg, pl; C, dir, obl; P, 1st, 2nd, 3rd stand for gender, number, case, person features.

            (Lexical Specification Rule 1)

            Det [Indef [k-] ] = k-o –t GnoC ‘ identity’

                                          k-o, k-u -i ‘ identity human’

                                          k-ay, k-ah-i ‘ identity non-human’

               any of these Duplicated ‘exhaustive, distributed, or graduated’ in addition.

            Where: k- ‘some … or other’


            Note that k-also marks the Interrogative in Marathi, while j-marks the Relative, t-the Distal Demonstrative, and h-/ y-/ the: Proximal Demonstrative.  (Compare Hindi which has j- Relative, k- Interrogative / Indefinite, y- Proximal Demonstrative, v- Distal Demonstrative, and s-/ t- Correlative.)  For the use of Marathi Proximal Demonstrative h-/y- as Correl see section III.


(Opd 1 continued)


            Conditions: (5) Rel and Correl are coreferential.  One cannot be human and

                                    The other non-human, for example.

                               (6) NP1 Residue and NP Residue are congruent in Gender, Number,

                                    and, if both N have N – stems, with identical N- Stems.

                               (7)  Sˊ is not either Interrogative or Imperative.




(Example 1) j-e      kah-i           (pustәk)…. t-e (pustәk)   ‘which (book)…

                               kay                                                            that (boek)’

                               ko -t- ә



(Ex.2)  ǰ -i         ko                  (ba-i) … t-i (ba-i ) ‘which (woman)…that (woman)’

                        ku -i

                        ko -t-i


(Ex 3) ǰ -i           t-i        (      ba-i       ) ...t-i          (   ba-i           )

                         mi             indir-a                     mi         indir-a

                         tu        r                       rˊ         tu        

            ‘which  that       (   woman   )...  that                (           woman    )

                         me            Indira                      me                    Indira

                         you     r                                   you       r                          rˊ


Where: r, ˊ = respectively

Compare ‘ǰ ‘ ci ‘bai’ /t/ (e) ti bai →’ ǰi’  ti ‘bai, also’ ǰi’ ǰi koṇi ‘bai ˊ tiˊ tiˊ bai

(Ex 4)  S1: mi                ko-tə              ja pustək        vik-ət ghet-l- ə

                   X1             Rel                   NP1 Residue     Y1


                    ‘I bought some fat book or other’


            S: ti-nə S1  t-ya             pustək-a       č -i    mag--i- ke-1-i

                 X            Correl NP/Residue      Y

                  ‘she asked for that book S1


            Where : ko -t- ə (like any of the Indefinite k- words) can be optionally

            Strengthened by the emphatic particle t- ər-i


(Ex 4a) Here are possible conversional exchanged that bring out the underlying



            Ex 4 illustrates a typical operand.  We are now ready to run through the family of transformations concerned with Relativization.


Relativization Transformation Family


We shall now take up the transformations one by one.  The serial alphanumeric labeling is for ease of reference and for bringing out the family relationships.  The order of application when it is not inherent in the rules concerned is made explicit under Conditions.  Obligatory singleton operations are necessarily meaning-preserving – more accurately, meaning-neutral.  Optional operations and the choice between alternate operations is naturally not so meaning-neutral.  The stylistic difference is indicated under Conditions where this could be done with some confidence.  obl  means ‘wherever the operand emerges’ and opt means ‘optionally where the operand emerges’ (obl of course standing for ‘oblique case’).  A semicolon stands for sentence-juncture. Opd 1 is presupposed in all grammatical trans-formations unless the contrary is indicated.


(Grammatical Transformation Rule 1) Relative – Correlative - marking

            (1a) obl NP1 [Det [=>NP1 [Det [ j-

            (1b) opt  j- k- k- => j- j- k-

            (1c) opt NPˊ [Det [t- => NP’ [Det [t- t- /NP1 [Det [j- j-


{ ǰekhi kahi] ambe əstil [te] kahun ghe: (what any  ) mangoes are, (these ) pickout.’

{ ǰekhi kahi] ambe] muhe əstil [teambe] kahun ghe (what any mangoes) large are, (those)

ǰe ǰe ambe əstil  te kahun ghe ==>ambe ambe əstil te.... ==> ambe ambe ...whatever mangoes those pickout

ǰe ǰe ambe mohe əstil te ambe kahun ghe ==> ambe mohe mohe teambe ...  əstil

ð                 mohe mohe ambe...whatever mangoes are those mangoes pickout.

ð                 (cf. ǰe   ambe mohe mohe əstil te ambeku hun ghe => mohe mohe ambe... what mangoes quite large those mangoes pick out.

Does the Relativizer make a difference to the meaning?  We shall take up this question in Section III.


(GTR 2) Relativizer Concord obl


              Opd : j-

              Operation: Copy out G, No, C from N under NP1 after j- unless nothing

              Intervenes j- and N [G, No, C]


(Ex 5)  j-e ko (-t- ə) ko -t-ə pustək … t-e pustək

            j-e j-e ko -t-e pustək … t-e (t-e) pustək


            ‘whichever book … that (each) book’ (pustək  being neuter, singular, direct).


            Actually GTR 2 is only a special case of a general rule governing any Declinable entering NP under the immediate domination of S, [AP], Det, or N.


(GTR 3) Extraposition in Matrix S opt


            (3a) Opd : S [X NP-Pc Y]

                                    1    2      3


            Where: Opd 1 is not presupposed; Pc = a postclitic if any (such as a postposition

                 Or an emphatic particle) taken out of Y


            Opn : (3a1) 2; 1 3

                      (3a2) 1 3; 2


            Conditions: (1) NP is not dominated by a higher NP

                         (2) GTR 3a1 is favored, if NP-Pc is longish or if it is the Given

                               element in the communication. If Opd 1 holds, GTR 3a is favoured  if Rel = Def or if X is likely to be mistaken for a part of X1

                                (say X1 being null).

(3)    GTR 3a2 is colloquial and is favoured only if NP is the New

        element in the communication.

(3b)  Opd 1 is presupposed.

         Opn : X NPˊ Pc Y; (ki) NP – Pc

Condition : S1 is an afterthought.


(GTR 4)   Extra-position in NP opt


                        Opd: NP [S1 NP’ [t-GnoC NP’ Residue ] ] Pc


                        Opn : (4a) NP [NP ˊ Pc(;) (ki) S1]

                                  (4b) NP [NP ˊ Residue Pc(;) (ki) S1 t-GNoC] Pc


                        Conditions: (1) GTR 4 is colloquial and is not common.

                                    (2) ; is favoured if NP ˊ is longish.


  (GTR 5)   Extraposition in Embedded S opt


                        S1 = > NP1 – Pc1 X1 Y1


                        Condition:  GTR 5 is obl if GTR 3b or 4 is being applied; it is favoured if GTR 3a is being applied or X1 is likely to be mistaken for a part of X (say, X being null).


            Note that GTR 5 is in a way an extension of GTR 3a1 to S1 but without the sentence juncture.


(GTR 6) Extra-position in NP1 opt


            Opd:   j  Rel – Def (N)

                       1       2          3 

            Opn:     (6a)  2 1 3

                        (6b)  2 3 1


(Ex 6) [ mi j-e ko -t- ə ja pustək vik- ət ghet-1- ə t-ya pustə k-a č -i ] ti-nə

            mag- -i ke-1-i ‘which fat book I bought for that book she asked’

            (GTR 3a1; putting [   ] last we get the result of GTR 3a2)


(Ex 7)   ti-nə t-ya pustək – a č-i mag- -i ke -1-i; (ki) ǰ -e ko -t- ə ja pustək

            mi vik- ət ghet-1- ə t-ya pustək-a č-i (GTR 3b, 5)


(Ex 8) ti-nə t-ya pustək-a č -i mag- -i ke-1-i; (ki) ǰ -a pustək-a  č -i mag-

            - -i t-i  kər- -ar n-ahi  əs - ə mə -la va -1- ə ho- t-ya’ she asked for

            that book which book I thought she wouldn’t ask for’ (where an

            S-node intervenes S1 and NP1)


(Ex 9)   h –e pustək [mi j-e k-o -t - ə ja pustək vik- ət ghet – 1- ə  t-e ] ah-e

            ‘this book is that which fat book I bought’.


(Ex10)  ǰ -i      t-i           ba-i ;        t-i            ǰ -i ba-i ;     t-i      ba-i  ǰ -i

                      mi                           mi                               mi


            (Cf. Ex 3; GTR 6a, 6b.)


            Ex 10 does not illustrate, as Junghare would presumably suppose, Antecedent-copying.  Her     examples (8) (ii), (8) (iii) also illustrate GTR 6, though she does not seem to be aware of the possibility of t- being a part of the Relative Clause also (Opd 1, Condition 1) contributing its distinctive meaning.


            The fulfillment of Opd 1, Conditions 1, 2 and the implementation of GTR 1, 3b, 4a, 6 result in various types of interrupted sequences of j- and t-. The implementation of GTR 1, Condition 2 and Opd 1, Condition 6 result in various types of interrupted sequences of N-stem under NP and N1-stem under NP.  The next GT Rule is concerned with the simplification of such pronoun and noun-stem sequences.


(GTR 7) Relative simplification opt


 (7a) Opd : S [S1 [j* ] t* ]

                              S [t* ; S1 [j* ] t* ] (cf. GTR 3b)

                              S [t* ; S1 [j* ] ]  (cf. GTR 4a)


            Where : the sequences may be interrupted.

                        j* = j or jt or tj or tj  (cf. GTR 1b, 6)

                        t* = t or it (cf. GTR 1c)

                        j   = j-GnoC and t = t-GNoC


            Opn :   (7a1) j* loses j or jj (not t) if t* does not precede

                        (7a2) t* is lost if j* follows and N in NP is not zero-stem.

                        (7a3) t* is lost if t* j* precede.


            Conditions: (1) GTR 7a1 is favoured if Rel = k-realized.

                                (2) GTR 7a2 is colloquial and is not common.


            (7b) Opd : S [N1 N ]

                              S [N N1 N]

                               S  [N N1]


            Opn: (7b1) delete one of two

                     (7b2) delete two of three


            Conditions: (1) Not applying GTR 7b1 is ponderous; not applying GTR

                                    7b2 even more so.

                         (2) In applying GTR 7b, don’t delete N or N1, if A precedes it

                               (except very colloquially).


            (Ex 11) ti-nə mi ja pustək vik-ət ghet-1-ət –ya-č-i mag- -i ke-1-i

                        ‘she asked for the fat book I bought’ (GTR 7b and Condition 2)

                        The absence of N-stems at all points is of course traceable to

                        GFR 1, Condition 2.





Relative Clause Abbreviation


            Abbreviating Relative Clauses becomes particularly urgent when there are more than one of them in the same NP.  In a sequence of the type NP [ S S S Det N] one carries out Relativization and its abbreviation beginning from the inmost S and working one’s way out to the highest NP, Relative Clause Abbreviation (GTR 8a, 8b) turns a Relative Clause (with a finite verb) into a Relative Phrase (without a finite verb).  GTR 8c fuses the Relative Phrase with the NP’.  GTR 8a below is the Marathi counterpart of the so-called Relative Participle construction in Dravidian languages.


(GTR 8) Relative Abbreviation opt


            Condition : Apply GTR 8 after GTR 1a, when greater compactness is feasible

            And desirable.

(8a) S1 [X1  NP1 Y1] = > S1 [AP[ Residue of X1 Y1 V Aux*] ]


 Where : Aux in the Operand and Aux* correspond in the following

               Manner (C stands for the Concord endings GNoP, NoP, GNo as the case

               May be)

                        Aux ending :                             Aux replacement

                        ah-C                                     - -ar-GNoC

                        -t-C (present)                         -l-el-GNoC


                        ho-t-C (past)                              -l-el-GHoC

                        -l-C (past)                                  -l-GNoC       

                        -t-C (present)                             -t-GNoC      

                        - -ar                                        - -ar


                        -av-C                                         -ay –c- GNoC





Conditions:       (1) NP1 is Subject, Object, Axis of postposition, the last

                  possibility being rather colloquial.

            (2) X1 Y1 is not inside an excessively long and complex sequence:

                  for example, NP1 is not inside S embedded in S1 as in Ex.8.

            (3) GTR3 has already been applied to any S more inward to NP

                  than the Operand S.

            (4) Aux* ending in –l-GNoC, -l-GNoC, - -ar are less common and

                              tend to be confined to set expressions (i.e., call for appropriate                        

                              LS rules).

                        (5) - -ar, - -ar-GNoC are not used if NP1 is not Subject, except

                              rather colloquially when V is intransitive and when certain

                              other conditions are fulfilled.

(6) –l-el-GNoC, -l-GNoC are not used if NP1 is not Subject V transitive, except rather colloquially in those cases where - -ar-GNoC is not       available.

(7)        –ay-c-GNoC is not used if NP1 is Subject and V transitive.



(Ex 12) mi vik-ət ghet-l-el- ə/ghy-ay-c-ə t-e pustək ‘that book bought / to be  bought by me’


(Ex 13) pustək vik-ət ghe--ar-ya / ghet-1-el-ya t-ya mə -la ‘to me who buys or will buy / has bought the book’


(Ex 14)             ǰ-a- č –i əš-i mət- ə ah-et t-o maus

                         əš-i mət-ə əs-l-el-a/əs-ṇ-ar-a-a t-o maus

                         ‘that man whose such opinions are’


(Ex 15) ǰ-a-t bhaǰ-i  mi-t-e  t-o bajar

                         bhaǰ -i mi - -ar-a bajar

                         ‘the market in which greens are available’;

                         bhaǰ-i mia-1-el-a bajar

                         ‘the market in which greens have been available’ ;

                         bhaǰ-i mi-ay-c–a bajar

                         ‘the market in which greens are to be available’


(Ex 16)             ǰ-a-la tap  a –1-a  ah-e t-o maus

                         tap a-1-el-a maus

                         ‘the man whom fever has affected’ ;

                         ǰ-a-la tap  ye-t-o  t-o maus ;

                         tap ye- -ar-a  maus

                         ‘the man whom fever affects'


(Ex 17) mə-la šikək  vh-ay-c-ə n-ahi; j-e ma-jh-e  vəil ah-et   ‘I won’t become a teacher, which my father is’ (cf. Opd 1, Condition 4 ; GTR 8a and GTR 8b too, are inapplicable, since NP1 šikək ‘teacher’ fails on GTR 8a, Condition 1 and the Opd for GTR 8b). (GTR 8 continued)

 (8b) S1  [NP1 Complement VAux] = > [S1] [                AP     [Complement*]   



Conditions: (1) V = əs / ah / ho ‘be’.

(2) Aux = Aux ending in ah-C, t-C (present), Future, ho-t-C  (past), -l-C (past) (where C stands for GNoP, NoP, GNo as the case may be), i.e., where GTR 8a can yield əs--ar-GNoC and əs-l-el-GNoC.

(3) When Complement is NP, then S1 = NP [Complement *] and S1 is called an Appositive Relative Phrase.  In GTR 8a and elsewhere in GTR 8b, S1 = AP and S1 is called an Attributive Relative Phrase.

(4) Complement = NP, AP, NP-obl-c-GNoU = Complement*

(5) Where Complement = AdvP, Complement * = AdvP-c-GNoC,

Where Complement –AdvP ending in at, baher, vər, puh-e,  mag-e, -kə-e, khal-i, -eth-e –ith-e/uh-e, ghər-i, məddh-ešev-i, etc., Complement* is Complement –il or more coll-oquially Complement-l-GNoU (-e- and –i- being deleted).

(6) GTR 8 has already been applied to any S more inward to NP

                        than the Operand S.

(7) When Complement is NP, GTR 8b is favoured with the following pairs of Complement and NP :


                                                Complement                             NP


                                                N-Proper                     N-Role/N-Namable

                                                N-Role                         N-proper

                                                N-Proper-Human-       N-Proper-Human-

                                                    given name                    surname


N-Role as Appositive to N-Proper sounds foreign (Perso--Arabic or English).  N-namable is a common noun like ghər-a-ə  ‘agnatic lineage’, nəd-i ‘river’, kutr –a ‘dog’, bha-a ‘language’, etc.


(8c) Opd : NP [S1 [ AP ]                  NP ’       [Det [t…] N]


                               Opn : NP’ [Det N[  AP     N] ]



Where: Det tends to lose –t/ əs - especially if it weakly demonstrative.



Conditions: (1) GTR 8c is favoured if AP is short, characterizing rather than antifying or identifying, quantifying rather than identifying; or if NP is short and simplex (i.e., with zero Det, singular number, and without embedded Relatives).


(2) If N has no N-stem in it and Det is zeroed there is no way to distinguish a fused Attributive Relative Phrase from an ordinary AP  (i.e., NPˊ   [Det [ʘ] N [AP] from AP) if the syntactic context admits NPs as well as Aps.  If N has no N-stem to start with, then there is no way to distinguish a fused Appositive Relative Phrase from an ordinary NP (i.e., NP’ [Det N [NP] ]  from NP’ [ N ] N-stem]])


(Ex 18) (j-e) səg-i-kə -un lal ah-e t-e (səphərčəndə) ucəl

                        səg -i-kə -un lal (əs-l-el- ə) (t-e) (səphərčənd ə) ucəl (GTR 8a,b)

            (t-e) səg-i-kə -un lal (əs-l-el- ə) (səphərčənd ə) ucəl (GTR 8c)

‘pick up the apple / the on which is red on all sides’ (if səphərčənd ‘apple’ is deleted, the ( ) around t-e ‘that’ are likely to operate in very colloquial speech;  if t-e be replaced by əs-ə it will be no longer NP-

Relativization, but AP – Relativization for which se Section III).


(Ex 19)             (j-o) ma-ih-a mitrə ah-e t-o ‘ who is my friend –he / that’

             ma-jh-a -mitrə (əs-ṇ-ar-a) t-o (GTR 8a, b)

            (t-o) ma –jh-a mitrə(əs-ar-a) t-o (GTR 8c)

                        ‘my friend; that friend of mine’


This last example of course shows where Noun-stems is Noun Phrases come from – they come from Complements in Relative Clauses.


(Ex 20)             t-i-c-ə a-i c-ə nir-əpekə

‘her mother’s unselfish love, her unselfish mother’s love’ (her ‘mother’s’, = ‘motherly’ in either case).


(Ex 21)             [čhətr-i la bəs- əv-1-el-i] [kimt-i] [pahr-i] [həsti-dənt-i]

muh  ‘the costly white ivory handle fitted to the umbrella’ (this is  [1][2] [3] [4] [N] ]]]] ; other orders such as 2 1 3 4, 1 2 4 3, etc., are possible but are marked, this being governed by considerations of style and situational context ;  another possibility is illustrated by [1[ [ 2 & 3 & 4]  [N]]]]  where: & stands for ai ‘and’ )


The ordering of “Relative S-nodes in the nesting or conjoining deep structure is fairly flexible in Marathi – the unmarked order and the degree of usability of the marked orders depending on considerations similar to GTR 8c, Condition 1 and the culture-bound preconceptions about intrinsic and extrinsic attributes.  S1 is not normally repeated in such nesting and conjoining;


(Ex.22)             pahr-i paṇḍhr-i muh ‘white white handle’ (this parallels not to:

                         muh pahr-i ah-e ‘the handle is white’ taken twice but to:

                         muh paḍhr-i pahr-i ah-e where Duplication shows ‘intensity’)


There are some Relative Abbreviations that are too idiosyncratic to be reasonably handled by GT Rule.  Lexical Transformations Rules will do the job.


            (Lexical Transformation Rule 1)

                        j-[GNoC] r  t-[GNoC ]r N-Stem P

                        j-[GNoC-P] r  t-[GNoC –P] r

                        < = j-[GNoC] r  ko -i ah-e/bhe -e-l/mi -e-l –[GNoC] r P

                        ‘no matter which’ (where: P stands for postposition but not an emphatic particle.)


(Ex. 17)            j-o (k-o -i bhe -e-l) t-o (maus)  t-ya-la sang- ət  ho-t-a

‘whomsoever he met with was telling him, i.e., everybody was telling   him’ (LTR 1 continued)


ǰ -a-c-GNoC t-GNoC where : the two GNoC concord ‘everyone as the case may be’


ǰ-a-c-GNoC t-GNoC where : the first GNoC concord with the implied

N = > ǰ-a-c-GNoC əs-e-l t-GNoC ‘whoever concerned’

nahi/nə -ko/va -e-l t-GNoC  ‘that which isn’t the case / isn’t wanted ; any that one may think of’


(LTR 2)


Proximal deictic with Time-unit-in-a-cycle where the deictic identifies the Time-unit with reference to the present moment can also be used by being referred to some moment removed from the present moment:


(Ex 24) aj ‘today’; aj<= aj j-o var ah-e t-ya var-i ‘on the same day of the week as today’ (similarly: ‘on the same day of the month/year as today’; similarly with: udya ‘tomorrow’ etc.)


Similarly with: Place-deictics.


(Ex 24)             aj t-ya-na su -i əs-t-e ‘on any day of the week that is the same as

 the day of the week today they have a holiday’ (cf. aj t-ya-na su -i

 ah-e ‘today they have a holiday’).


(LTR 3)

                        N1 N2 <= N2 -obl-c-GNoC N1

                        Where: N* and N* are from the following favoured pairs:


                            N1                                                                N2


N-Proper                                 ghər-a- ə ‘agnatic lineage’

                        N-Proper-Human-                   N-Proper-Human of N1’s family on

                             Given name                   father

                        mas-a ‘fish,                              Name of variety of fish, rice, etc.

                        tandu ‘rice’, etc.                                respectively



(Ex 25)             dad-o b-a paḍurəŋg

                         paḍurəŋg g-a c-a dadoba

                        ‘Dadoba, son of Pandurang’;

                         dad-o b-a tərhə -kər

                         tərkhə -kər-a-nc-a dado –ba

                        ‘Dadoba of the Tarkhadkara’;

                         Dadoba of the Tarkhadkars’;

                        dad –oba paḍurəŋgə tərkhə -kər

                         (full name of the well-known grammarian of Marathi)


            (Ex 26) rohu mas-a

                         rohu c-a mas-a

                         ‘rohu fish’ (but not pimpəl jha by the side of  pimp½-a c-ə jha and

                         pimpəl both meaning ‘a papal tree’)





Extrapolation by Restatement


            The correct interpretation of coreferentiality in the following example (Opd 1, Condition 5,6) forces us to recognize AP- Relativization as distinct from NP-Relativization.


(Ex 27) ǰ -it-k-e k-iti  gəhu  t-it-k-e tandu ½  ‘as much wheat so much rice, i.e., rice

of that same quantity of which quantity the wheat …’ (coreferentiality

holds, inspite of ‘wheat’ being different from ‘rice’, between the

respective quantity indications)


            The paraphrase under Ex 27, however, also suggests that AP-relativization is at bottom not different from the NP-relativization that GTR 1-8 deal with.  Actually GTR 3a is quite general not only in the sense that it can be applied to any NP whether it has a relativizing S in it or not, but also in the sense that it is quite parallel to AP-Extraposition.  These remarks also apply to AdvP-Relativization and Intensifier P-Relativization.



(Opd 1 Restatement)


            Replace NP, NP΄, NP1respectively by FP, FP΄, FP1 where: FP (Focalizable Phrase) = NP, AP, AdvP, IntensifierP, Rel and Correl will be redefined in the appropriate manner so as to accommodate:


            AP΄  [Det  [Indefinite] A [Quality / Quantity / Size / Ordinality]]

            k- əs (l), k-it (k) -, k-evh-, k-it-v-respectively

            AdvP΄  [Det  [Indefinite]  Adv [Place/Time/Degree/other-Circum/etc.]

            k-uh-e, k-evha, k-itpət, k-əs-, etc.  

            AdvP΄  [Det  [Definite]  Adv  [Time –to / Place /etc.]]        


            AP΄ [[ Det [Indefinite ]  Intensifier]  [Det A]]

and the like.  Proximal h-/y-/ are more freely used under AP Correl and AdvP Correl.


            GTR 1-8 will also call for suitable adaptations and the setting up of conditions to accommodate for gaps in the isomorphism.


            All this of course leads to recognize a parallelism in the constituent structure of Focalizable Phrases.



(GFR 2) FP Specification


            FP        Det F

                           S FP


            Where:  F = N, A, Adv, Intensifier (but not V ! )


            Apart from the adjustments in the Relativization Transformation Family, some LT Rules similar to LTR 1-3 will have to be set up to account for abbreviations like:


(Ex 28) (i)         j-əs-a t-əs-a ‘so so’

                        (ii)        i-ith-ə-a ;t-ith-e ‘comme il faut’

(iii)               t-o j-əs –a l-ah-i ma-jh-a mul-g-a (əs-av-a t-əs-a) ah-e

‘ he is, as it were, my son’


            Confusions between different types of Relativizations must be avoided.  The Rel and Correl markings may mismatch or be deceptive in other ways. 


(Ex 29)             j-e-k-iti gəhu ah-e-t t-e (gəhu) dy-a  ‘what little wheat there is – give it’;  this may be realized as:  j-it-k-e gəhu   ah-et  t-e (gəhu)  dy-a  ‘as much wheat there is – give it’


Ex 29   still exhibits NP-realization, in spite of the surface collapse of j-k-A Base to j-Abase.


(Ex 30) ǰ -i- č -at  t-o  bəs-l-a  ho-t-a-t-i  khik-i ‘the window in which he was seated’; alternately;  ǰ-ith-e-t-o bəs-l-a ho-t-a  t-i  khik-i, ‘the window where he was seated’


Ex 30   still exhibits NP-realization, in spite of the surface collapse of j-GNoC postposition to j-AdvBase.


(Ex 31) ǰ-it-k-e  k-iti gəhu  ah-e-t  t-it-k-e- (gəhu)  dy-a   ‘as much wheat there is – give so much’;  alternately j-e k-iti  gəhu  ah-e-t  t-it-k-e (gəhu)  dy-a  ‘what wheat there is – give as much’


Ex 31 still exhibits AP-relativization, in spite of the surface collapse of j-Abase-GNoC to j-GNoC.


(Ex 32)  j-o mn-a-t-l-ə ləp-əv-un thev-t n-ahi s-a maus; mə-la kal bhe-l-a’

              I met yesterday a man of the sort that doesn’t hide what’s in his mind’


            Ex 32 is, like Ex 31, one of AP-Relativization except that j-o cannot be ‘restored’ to j-əs-l-a without changing sense.


(Ex 33)             ma-ǰh-i bəh –i ǰ-i ek-da nigh-un ge-l-i t-i punha a-l-i n-ahi

                        ‘once my sister was gone, she never came again’;  alternately:

                        t-i  punha a-l-i n-ahi; ma-jh-i bəh-i ǰ-i ek-da nigh –un ge-l-i t-i


            The alternate version with its two occurrence of t-I (first as ProNoun, second as declinable ProAdv) brings out that Ex 33 exhibits AdvP – Relativization, j- and t- acting adverbially.


            Ex 34 exhibits not AP-Relativization but IntensifierP – Relativization.


(Ex 34)             galič-a [khol-i ǰ-it-k-i k-ah-i lamb ah-e t-it-k-ya peka k-ah-i lamb] ah-e’ the carpet is somewhat long in relation to that extent to which extent the room is long’; more compactly:


galič-a [khol-i ǰ-it –it-k-i lamb ah-e t-ya –peka lamb] ah-e;

even more compactly : galič-a [khol-i peka lamb]ah-e

The respective bracketed portions of Ex 34 have the structure:

AP [IP1  [PP  [IP]  [S1I2 ]  Postp]  I1 ]  A]


Where : (1) IP : Intensifier Phrase, I: Intensifier, PP: Postpositional



(2) S1 = [khol-i[ǰ-it-ki] lamb ah-e] with Rel-I3= ǰ-it-k-i

I2 = t-it-k-ya = t-ya = (Correl)

Postp = peka ‘in relation to’

I1 = k-ah-i  ‘somewhat’ =


It will be noticed that I1 does not mean ‘more’.  I1 could have been ədhik ‘more’ with the same result or kəmi ‘less’ with the opposite result.  Also, IP2 could have been tacked straight on to A without using Postp or I1.


(Ex 35) galič-a [khol-i ǰ-it-k-i lamb ah-et-it-k-a lamb] ah- e’ ‘the carpet is long to that extent to which extent the room is long’; more compactly :  galič-a khol-i ǰ-it-k k-a lamb] ah-e


It should be obvious that the more extreme compressions shown under Ex 34 and Ex 35 would be impossible if S1 were even a slightly more elaborate affair, say:


ǰ-it-k-i  lamb khol-i  mi½-a-l-i

            ‘the room long-to-which-extent was available’


            On the other hand the more extreme compressions would be forced on us if the whole of Ex 34 or Ex 35 were to act as Relative Clause to t-i   khol-i so as to yield the meaning ‘that room than / like which the carpet was more / as long’.


            Comparisons of Degree involve us in IP-Relativization.


            We just cannot have Postp-Relativization to Marathi.  The best we can do is a paraphrase.  Suppose the brilliant detective cannot pin down whether the criminal put the bottle on or under the chair yesterday.  All that he can say with confidence is the following:


(Ex 36) S1: t-ya-nə aj ebl-a-č-a [k-uh-e təri] kəp hev-l-a

                        S: t-ya-nə kal khurč-i-č -a [S1] t-ith-e-c] bal-i hev-l-i

            Thus we get:

t-ya nə kal khurči-č-a [[t-ya-nə aj ebl-a_č-a ǰ-ith–e kuh-e təri kəp t hev-la] t-ith-e-c] bal-i hev-l-i

                        ‘he put the bottle there around the chair yesterday where around the table he put the cup today’


Ex 36   actually exhibits AdvP –Relativization.


            Nor can we have Verb Relativization in Marathi.


(Ex 37) t-ya –n ti-la-i-e ke-l-ə te-e mi tu-la ke-l-ə

                        ‘I did to you just what he did to her’

Ex 37   actually exhibits NP-Relativization.


            Can the Sentence as a whole take on a Relative S ?  The answer to this is, perhaps surprisingly, yes.  But this cannot be fitted into GFR 2.


(GFR 3) Circumstantial Insertion


            S    Circum  [S1] S΄

            Where : (1) S1 = [Complement [AdvP]  NP [Det  [Indef] N [xistable / Happenable ]]   V  [ Exist / Happen ]  Aux].


            (2) NP and S1 [are coreferential.


                 This triggers a transformation.


(GTR 9)  Circumstantial Assimilation, obl


            Opd: S [Circum [S1]     [X Y ] ;  Opd 1 is not presupposed

            Where : X is the first constituent (normally the subject) dominated by S;

            Y is the residue which may have previously inserted Circum at the beginning;

            S is specified by GFR 3.

            opn: S [S΄{X Circum [AdvP ] Y]


(Ex 38) S1 : bag-e-t kahi jh-a-l-ə ‘something happened in the garden’.


            S: [S1] t-ya –nə nəntər layər khiš -a-t  hev-l-a ‘[S1]

[S1]  he later put the lighter in the pocket’


            After GTR 9: t-ya-nə   bag-e-t nəntər layər khiš -a-t hev-l-a

            ‘in the garden he later put the lighter in the pocket’


            Note the difference between English and Marathi.  Relative Clauses and Circumstantial ‘normally’ follow their heads in English; in Marathi the Relative clause precedes its head and the Circumstantial precedes the bulk (i.e., Y in GTR 9) of its head.


(Ex 39) t-ya-nə nəntər bag-e-t layar khiš -a-t hev-l-a


            ‘Later in the garden he put the lighter in the pocket’


            Note the difference in the status of AdvP bag-e-t ‘in the garden’ as Circum and AdvP khiš -a-t ‘in the pocket’ as one of the Complements of the V hev ‘put’.  GFR 3 permits nesting-the more intrinsic the Circumstance the more inward it is normally placed.  (Compare the order in Ex 38 and Ex 39).


            GFR 3 is probably a special case of a more general rule in which Neg and other elements figure in place of Circum.


            The next extrapolation to be considered is the possibility of Multiple Relativization. 


(Opd 1 Restatement continued)


            There may be more than one FP under S and correspondingly more than one FP1 in S1.  That is to say, Operand S may be a transform of the following sort:






            S*                   S*                                                                     S1.... FP*    FP**...




X *   FP *    Y*  X**    FP**  Y**


S1                FP΄*         S1**     FP΄**




X1    FP1*       Y1  *     X1 ** FP1  Y1  **


Where S* = S**;  S1* = S**;  on the other hand :  FP** FP** ; 

FP1* FP1**  (Ex 40)


            S1 : ko-i kah-i  iččh-i-1  ‘someone or other will desire something or other’

            S : [S1 ] t-ya-la  [S1 ]  t-e-mi½-u de   ‘let him get it’


            After GTR la : j-o ko-i  j-e kah-i iččh-i-l ;  t-ya-la  t-e mi½-u  de


            ‘whoever whatever desires, let him get it’  (the English gloss, it may be

            noted,  turns S1 into a Circumstantial of the sort specified by GFR 3 and

            GTR 9)


(Ex 41) ǰ-a mas – a  la  j-e-bəkis  mi½-a-l-ə  ho-t-i  t-o  maus t-ya  bəkis-a

             vər khu ho-t-a   ‘the man who got the prize was happy to get the prize

             he got’.


            The gloss for Ex 41 in English, which does not permit Multiple Relativization, of course recalls the Emmon Bach Conundrum.  A more literal though un-English gloss would be  ‘Which man which prize got that man was happy with that prize’.


            The extrapolation now to be considered takes out a skeleton of long standing out of the linguist’s cupboard: the so-called Sentential Complementation.  To understand its nature we have to go back to GFR2.  Normally S1 and S residue can change roles with suitable adaptation of Det of the respective focalizable phrases.


(Ex 42)   mi ti-nə ǰ-a ko-t-ya pustək-a č-i mag--i ke-1-i te ja pustək vik-ət ghet-lə   ‘I bought that book which she asked for’


            This is of course by no means a paraphrase of Ex 6; though it could but needn’t be referring to the same events.  The reasons for this difficulty will become apparent later.  Such a difficulty will not arise if both Focalizable Phrases are carrying Det *Definite*.


(Ex 43) ǰ-i/t-i mi ba-i roj ah tas kam kər-ay-č-i t-i/ǰ-i mi ba-i aj don tas-a-t        dəm-un jat-e   ‘I who used to work for eight hours a day am tired out in two hours today (ǰi….ti) ;  ‘I who am tired out today in two hours used to work for eight hours a day (ti….ji)’


            Now can we look for such reversion of roles between S1 and the residual S in GFR 3?  Compare Ex 44 with Ex 38 keeping this in mind.


(Ex 44)   [t-ya-nə nəntər layər khiš a-t  hev-1-a;]  t-e bag-e-t  jh-a-l-ə] ‘he later put the lighter in the pocket;  that happened in the garden’;  OR‘ that he later put the lighter in the pocket happened in the garden’


(Ex 45)    [t-ya-nə nəntər layəer  khiš-a-t  hev-1-a;]  j-e bag-e-t  jh-a-l- ə ‘he later put the lighter in the pocket, which happened in the garden’;


            The first-mentioned gloss for Ex 44 is of course not under consideration here; incidentally it calls for a different sentence-tone than the one we normally have for extra-posed phrases put at the beginning.  Ex 44 under the second interpretation has a front-extra-posed abbreviated Relative Clause.  Compare Ex 45 with Ex 44 and Ex 38.  Ex 45 is closer in meaning to Ex 38.  But in respect of form Ex 45 is closer to Ex 44, in that, in both Ex 44 and Ex 45, the bracketed clause is appositive to a Noun Phrase (t-e and j-e respectively).  The apposition is essentially of the sort described in GTR 8b, especially Condition 7.


            The following version of Ex 44 is unacceptable which would keep the Relative Clause intact.


(Ex 44, hypothetical version) * [ [j-e [ t-ya-nə nəntər  khiš -a-t  hev-l-a]  ah-e] t-e]

bag-e-t jh-a-l- ə,


Ex 45 is actually a Minor Sentence in which an NP stands by itself as a Sentence:


(Ex 45, Structure) S [ NP [ NP ΄ [ S]] NP ˝΄] ;  S1 [ j-NP1 X1 Y1 ˝] (where

            GTR 4 has been obligatory applied to NP and NP1 could be t-e or zero)

            In structures of this type S1 is usually, perhaps always, ‘non-restrictive’.


(GTR 4, Restatement)


            Add one more Condition : Normally opt, but obl if NP ΄ = [ S1 [ NP [S] ] NP˝]

where j- has been inserted in S1


            Before we tackle the main question, let us go back to GTR 8c, Condition 2 which seems to promise that we can characterize all occurrences of N-stems inside Noun Phrases as appositives.  In fact this has been seriously proposed.  But this proposal as well as the similar proposal to introduce all Sentences under NP (other than the Relatives of course) as appositives land us in infinite regress.  What we need is a restatement of Noun Specification.


(GFR 1, Restatement)


            (1a)          Det N

                                S   NP


            Condition: If S is present, the operand NP conforms to Opd 1.

            (1b) Det    Indefinite / Definite, Relative / Interrogative / … (Det-stem)

            Condition: Det = Det-stem obligatorily if N = S.

            (1c) N    Gender, Number, Case, Existable / Happenable, Human / Non-

                               human, …. (N-stem /S)

            Condition: N = [ N-stem/S ] only if the Operand NP appears in the

            Environment S[ -V [ Exist / Happen …] Aux].  Typical verbs are  əs/ah/ho

            ‘exist’, ho/jha  ‘happen’, ghə  ‘happen’.


(GTR 10) Distancing of N

            (10a)    Opd : S [ …NP […N ]  …] Opd 1 is not presupposed

                        Opn : S [ …NP […]  …]  ; ki [ N]   (Viz)

            (10b)    Opd:  S [ …FP […F ]  …]

                        Opn:  S [ …FP [[…F ] Viz ] …]

                        Where : Viz = mhə - ǰe, mhә -un

            Conditions for (10a, b) : (1) Both bring about a certain distancing – it is

            As if the listener is invited to look at N.

(2) If F = N = [N-Stem], GTR 10a doesn’t apply; GTR 10b is opt; it is

favoured if V-Aux = əs-t-GNoP.

            (3) If F = N = [S] , GTR 10 is obl; either 10a or 10b applies.  However,

            GTR 10 does not apply if GTR 8d follows.  (GTR 8d is yet to be stated.)

(4)Conditions 2, 3 presuppose that S is specified by GFR 1c, restated,

            Condition.  Elsewhere other conditions apply.


            Now let us look at some typical sentences in which N [N-stem] and N [S] are permitted in the deep structures of the sort visualized in CSR 1c, Condition.


(Ex 46) t-e pustək ho-t-i   t-i-nə  t-e magit-l-i   ‘There was that book.  She asked for it’.


(Ex 47) k-o-t-ə ek pustək ho-t-ə  t-i-nə t-e magit-l- ə.  ‘There was a book.  She asked

            for it.


(Ex 48)             k-ah-i ek pustək mhə-un  əs-t-ə   t-i roj t-e ek mag-t-e.   ‘There is such a thing as a book.  She asks for one daily.’


(Ex 49) k-ah-I ek pustək mhə-un əs-t-ə j-e ko-t-ə pustək  əs-t-e t-e-t-i-la    avə-t-ə.  ‘There is such a thing as a book.  Whatever is a book she likes it’.


            These are examples of N [N-stem] sentences – respectively serving as starting points of Definite Specific, Indefinite Specific, Indefinite Non-Specific Non-Generic, and Indefinite Non-Specific Generic reference.

            Now for some N [S] sentences.

(Ex 50) [ t-o cuk –1-a] mhə -un k-ah-i jhal-ə

             əs-ə kah-i jha-l ə  ki [t-o  cuk-1-a]

            ‘something happened – namely, he erred’


(Ex 51) [ t-o cuk –t-o] mhə -un k-ah-i  ho-t- ə

             əs- ə kah-i ho-t-ə  ki [t-o  cuk-t-o]

             ‘such a thing happens –namely, he errs’


            These then are the atomic propositions so to say that constitute the referential infrastructure of a discourse.  They necessitate the awkward structures NP [Det N [Nstem] ] and NP [Det N [S] ] – precisely because this is the way to eliminate them from the rest of language.  The skeleton in the cupboard turns out (pardon the mixing of metaphors) to be skeletal key, after all.


            On the other hand, N and S enter only through appositives elsewhere.  We must resist the temptation of recognizing NP [S] and NP [S N] on the same footing.  The problems raised by Factivity have to be tackled by frontal attack.  They cannot be ‘fixed’ by doctoring the NP specification Rule as Kiparsky does.  Parenthetically, I may suggest that GFR 1c may have to provide for, say,


            N [ Fact / Nonfact / Counterfact]

In the same way we have already provided for

            N [Existable / Happenable]


            All such features will be relevant when coreferentiality is defined for the NP1 of atomic propositions and the NP΄ to which they are in apposition.


            Here are some example of NP [S1 NP [Det N]] where S1 is a relativizable atomic proposition.


(Ex 52) j-e k –o -t- ə ek pust ə k ho-t- ə t-e’  ‘what book existed – that’

            pustək əs--ar-ə / əs-l-e-l-ət-e  ‘book-being that’ (GTR 8a)

            pustək t-e ‘book that’ (GTR 8b)

            t-e pustək  əs--ar- ə / əs-l-el- ə  ‘that book be-er’ (GTR 8a, 8c)

            t-e pustək  ‘that book’ (GTR 8b, 8c)


(Ex 53) j-e h-e jha-l-ə t-e ….;  ki t-o  cuk-l-a ‘what happened – that …; that he erred’

            (GTR 10a)

            jha-l-el-ə t-e; ki t-o cuk-l-a ‘happen-er that; what he erred’ (GTR 8a, 10a)

            t-o cuk-la-t-e ‘he erred-that’ (GTR 8b)

            t-e jha-l-el- ə …;  ki t-o cuk-l-a ‘tat happen-er;  that he erred (GTR 8a, 8c, 10a)

            te…; ki t-0 cuk-l-a ‘that…; that he erred’ (GTR 8b, 8c, 10a)


            (Note that GTR 10 is obligatory at some points, only GTR 10a is illustrated here.)


            It will be seen that such appositives are subject to GTR 8b, which needs to be restated.


            Further, sentential appositives are also subject to another abbreviation.  (GTR 8, Restatement)


            Let 8a be relabeled 8a1 and 8b be relabeled 8b1.  Then add:

            Opd: NP [S1 [NP1 [ Det N] V Aux ] NP ΄]

            Opn: (8a2) NP [S1 [ AP [ N V- -ar-/l-el-GNoC] ] NP ΄]

                     (8b2) NP [S1 [ NP [ N ] ]  NP ΄]


            Conditions: (1) N = N-stem/S; if N = N-stem, : NP ΄ = Det N []

            (2) V = [Exist / Happen]; e.g., əs/ah/ho  ‘exist’  ho/jha  ‘happen’,

            ghəd ‘happen’

            (3) Not applying GTR 8a2, 8b2 is ponderous;  8a2 is more ponderous

            than 8b2.


(8d)      Opd: NP [ S1[ NP [S ΄  [ …Aux ] NP ΄ ] ] ] ; GTR 10 hasn’t been applied to S.

Opn: NP [ S1[ NP [S ΄  [ …Aux *] NP ΄ [] ] ] ]

            Where: Aux and Aux* correlate in the following manner:



Aux Containing Aux*                Context


                                    -t-                    -t-e                   - ve ½ -i

                                    -av-                  -a-y-                 -la,c-

                                    -l-                     -l-ya-                -nəntər, vər, c--

                                    -av-                  - -n-sg-C       -vər, sah-i, etc.                      

                                                                                    also with direct Case

                                    any                   -i/e-                  -pəryyənt ə (s)to

                                    any                   zero



            Condition: If Aux* = zero, then NP ΄ = [əs-GNoC] and not zero and the

            Subject under S ΄ is obligatorily deleted.

            When any sentence appears in S1, as coreferential to NP, the Correl can be y/h- or əs- or by the side of t-.  So GTR 10, needs to be restated.


(GTR 1, Restatement)

            Add (1d):

            (1d1) t- = > h/y- / əs- /

            Conditions:  (1) S1 = [NP1 [Det N  [S1 ] ] V Aux]

            (2) Obligatorily t = > if S1 has undergone GTR 10b, and 8b2 (so that NP

                   ends up as NP [S ΄ Viz] )

            (3) Optionally t = > h/y- / əs- elsewhere; unless GTR 8d has been applied.


            Conditions governing the choice between mhə -un, mhə - ǰe, t-e, h-e, *s- and the use of ki need to be investigated.


(Ex 54) [t-o/t-ya-nə  cuk-t-e] ve½ -i  ‘at the time of him erring’

             […. Cuk-a-y-] la   ‘for him to err’

             [….cuk-l-y-a] nəntər  ‘after him erring’

             [… cuk- -ə]  ‘he being in error’

             […. –e-]pəryyəntə  ‘up to him erring, until he errs / erred’



            Compare: t-ya-c-ə cuk--ə ‘his being in in error’


(Ex 55)  i-ke bəs; t-i kḍe  ut; s-t-ya-nə ke-l- ə

            ‘sit here, get up there- so he did’


(Ex 56) [t-o cuk-l-ya ] c- ə mə -la hauk n-vh –t- ə

            ‘I didn’t know of him being in error’


            An idiomatic adjustment has caused some confusion in the discussion of Sentential Appositives in the past.


(GTR 11) –c-deletion opt


            Opd: NP [S1 [AP [NP*-obl-c-GNoC ] Det N ] ]

            Opn: obl-c = > dir

            Conditions: (1) Apply when N = kəlpəna ‘idea’, təkrar ‘complaint’,  həkigət

‘account’,  šəŋk-a  ‘doubt’ but not when N = karə ‘cause, reason’,

                                    ci ‘resentment’, etc.

                               (2) NP* = [NP [S] NP [ h-/t-] ]  or = [ h-/t].


(Ex 57) (t-o cuk-el) h-ya- č -i/h-i kəlpəna mə-la ho-t-I  ‘I had a notion of that /that

             he will err;  I had a notion of that / I had that notion’


            Sentential Appositives are subject to one more rather peculiar Relative Abbreviation involving AdvP.


            First, we need a GF Rule to bring about the following pairs (which are not like, say ǰ-evha…..t-evha  ‘when….then’) :


            j- ər ….t- ər ‘if ….then’

            j- ər -i….t- ər -i   ‘though…yet’

            j- əu (k-ah-i) …. əs- ə  ‘as if so’

            j- əs- ə (k-ah-i) … əs- ə  ‘as if ….so’


(GFR 4) AdvP Specifications (among other things; compare GFR2)

             AdvP 8 NP-P [NP [ S NP ΄ [ Det N] ] P]

(Ex 58)  t-o cuk-1-a  h-ya kar -a  sah-i  ‘for the reason that he erred’


            Now, the sequence [NP ΄ [Det N] ] P may be represented lexically be certain Adverbs.  But the Relative S is attached not to Adv, but to the latent N.


(GFR 4 continued)


            Adv P NP [S] Adv


            Where: Adv = mhə ‘for that purpose or reason, therefore’, mhəṇ- ǰe

            ‘under that condition’, ki  ‘under that condition’,  t-ər  ‘under that condition’,

            t-ər‘even under that condition, yet’,  əs‘in the manner recalling it’

            (the last one also usable adjectivally)


(GTR 12) Appositive Relative Marking opt


            Opd : AdvP [NP [S] Adv], OR : AP [NP [S] A]

            Where : Adv = t- ər   A = t- ə r-i əs-; A= əs-

Opn : Insert correspondingly j- ə r, j- ə r- i, j-əṇu (k-ah-i), j-əṇu (k-ah-i)

            at the beginning of S.  For əs-, alternatively, insert j-ə-s-

Condition: The resulting pairs are then subject to Relative Simplification

            (GTR7) as also Relative Extrapositions (GTR 3b, 4, 5, 6) within certain limits.


(Ex 59)[j- ər/j-r-i/ j- əṇu (k-ah-i)/ j- əs –a (k-ah-i) t-o cuk-l-a] t- ər/ t- ər-i/ s-/ s-‘if he errs then, though he errs yet, as if he erred’.


            Relative simplification of j-əs-a k-ah-i ….t-əs-a may lead to confusion.

Actually there are five similar-looking but distinct constructions with əs-, the first three  being Appositives and the other two being the Standard Relatives.


(Ex 60) (i) [AdvP [NP [S1] Adv [əs-] ]   ] by GFR 4, GTR 12

             (j-əs-ə ) (k-ah-i) mə-la kə½ - ət- əc  n-ahi (ə) s-ə tu bol-t-o ah-e-s

             ‘as if I don’t understand at all – so you’re speaking’

            (ii)  [AP [NP [S1] A [əs-]  ]   ] by GFR 4, GTR 12

            (j-s- ə) (k-ah-I) mə -la kə½ - ət- əc n-ahi (ə)s- ə m ə -la jha-l- ə

            ‘as if I don’t understand at all – so I felt’

            (iii) [NP [NP [S1] N []*s-] ] by GTR 8b, Sentential Appositive

            m-la kə½ - ət- əc n-ahi (ə) s- ə /h-e pah-un t-o həs-l-a

            ‘seeing that I don’t understand things, he laughed’

            (iv) [AdvP [S1 AdvP [Adv [əs-] ]  ] ]  by GFR 2, AdvP-Relativization,

            Relative Simplification

            (j-s-ə) mə-la kə½-ət- əc n-ahi t- ə s / əs- / əs- tu-la hi kə½ət n-ahi

            ‘just as I don’t understand things, so you don’t understand’


            (v) [AP [S1AP ΄ [əs-] ] ] by GFR 2, AP –Relativization, Relative


            (j-əs-ə) kam mə-la ye-t-ə t-ə s- ə / s- kam t-ya-la hi ye-t-ə

            ‘what sort of work I can do such work he can do too’


            Note that əs- can be colloquially shortened to –s- in (i), (ii), (iii), especially in certain set of phrases; that əs- ə is much rarer that t- əs- ə in (iv) and (v) except in the so-called Non-Restrictive Relativization in cases of the sort illustrated by Ex 32, and in Attributive Relative Phrases; that under (iii) əs-ə and h-e tend to be differentiated:  əs-ə for N [Nonfact / Fact] and h-e for N [Fact]; AP under (ii) and (v) is open to nominalization through GTR 8c.  Concordwise there are two competing patterns: (a) Suspended Concord (i.e., neuter singular) for all five;  (b) Adjectival Concord with an N for (i), (ii), (iii), (iv), (v).


            Constructions (i), (ii) may be used very loosely -əs- acting almost as a recapitulative; as in : dis-e-na-s-a ‘couldn’t -be –seen + as’ (i.e. invisible) həv-a-s-a ‘is-wanted + as (i.e. desirable, that which can be desired)’, vər-č-a- patə½i vər-un əs-a   ‘on a high level + as (i.e. as if on a high level, on a high level, so to say)’.  Consider also (iii) as in tumh-I ge-l-a-t-s-e?  ‘how come you went?’ or how come you went?  And tumh-I-s-e ge-l-a-t? ‘how come you went?’ involving a LTR abbreviating S əs- ə ka?  ‘S-why so?’


            Before we take leave of Sentential Appositives there is the possibility to be considered that the appositives, S with h-, t-, əs, and Viz may be Interrogative or Imperative as much as Assertive.  (This is of course not inconsistent with Opd 1, Condition 7, since S appositive is not the same as S1 but dominated by S1.)  By the side of Ex 61, we have Ex 62, 63, 64 also.


(Ex 61)             t-o cuk-1-a h-e mə -lapə l-ə

h-e mə -la pə -l-ə; ki t-o cuk –1-a (mhə -un)

t-o cuk –la mhə -un mə -la pə -l-ə

                        ‘that he erred I came to accept;’


(Ex 62)             t-o cuk-l-a ka əš-i šəŋk-a; mə -la a-l-i

əš -i- šəŋk-a mə-la a-l –i; t-o cuk-l-a ka (mhə -un)

                        t-o cuk-l-a ka mhə -unšəŋk-a; mə -la a-l -i

                        ‘whether he erred I came to doubt (of)’


(Ex 63) k-o cuk-l-a h-e/ t-e mə -la hauk n-ahi

h-e mə -la hauk n-ahi; k-o cuk-la t-e

                        ‘who erred this / that I don’t know’.


(Ex 64)             tu cuk –u nko-s əs- ət-ya-nə mə -la saŋgit-l- ə

əs-ə t-ya –nə mə -la sangit-l- ə; ki tu cuk –u nəko-s (mhə -ucn

                        t-u uk-u nəko-s mhə -un t-ya –nə mə -la saŋgit-l- ə

‘don’t you err-so he told me’


            Note that GTR 10 will need a slight restatement.


(GTR 10, Restatement)

            Add to Condition (3): If [S] is a Lacunal Question then Viz cannot be

            Used with it, thus eliminating one alternative of 10a and the whole of 10b.

            Ex 63 is of course distinct from Ex 65


(Ex 65)  k-o cuk-l-a t-o puh-e ye-i-l

            ‘who erred will come forward’ (j-o deleted before k- o by GTR 7a1.)

            And Ex 62 is distinct from Ex 66, which is exactly like Ex 61.


(Ex 66)             t-o cuk-l-a əš -i mə -la šəŋk-a; mə -la a-l-i

əš-i šəŋk-a mə-la a-l-i; ki t-o cuk-l-a mhə-un

t-o cuk-l-a mhə un šəŋk-a; mə-la a-l-i

                        ‘that he erred I came to guess’


            Furthermore Ex 62 but not Ex 66 shows c-deletion (GTR 11).  əs-i can be restored to h-ya-č-i ‘of this’ in Ex 62 but not in Ex 66.


            Polarity question (as in Ex 62) and Lacunal questions (as in Ex 63) appearing as sentential appositives have two interpretations each –N [Question] (as in Ex 62) and N [Answer] (as in Ex 63).


            Finally, we come to the so-called Non-Restrictive Relative Clauses.  The problem afflicts not only the Relative Clauses attached to NP but also their abbreviated versions yielding Attributive Adjectives and Appositive Nouns (what I have called Relative Phrases).


(Ex 67) ma-jh-a mha-tar-a bap  ‘my old father’


(Ex 68) bharət-a č-a məha-mantri indir-a-gandhi  ‘Indira Gandhi, India’s Prime Minister’

            (Appositive Relative Phrase)


(Ex 69) h-e ləhr-I gay-ək ‘these singers full of whims’ (usual interpretation Non-Restrictive, given the stereotypes about singers among Marathi speakers!)


            It also afflicts Relative Clauses attached to AP, AdvP, Intensifier P and their respective abbreviations, and also Circumstantials (inserted by GFR3).


(Ex 70)             t-o ek- še əynši senṭi-miṭər uncə ah-e; ǰ-itk-a uncə t-ya –a thor –l-a bhau suddha n-ahi  ‘he is 180 centimetres tall, to-which degree even his elder brother is not tall’.


            (This incidentally strengthens our proposal for GFR 2.)


            Junghare mentions the deletion of Correl as a distinctive feature of the Nonrestrictive.  This is less than a half-truth.  In the majority of cases (consider Ex 69) there is no overt difference; sentence-tone needs to be investigated but is not likely to amount to much.  The only markers are:


(i)                  Non-Restrictive favours Extraposition of Re1 after Correl (GTR 3b, 4, 6); loss of t by GTR 7a3 (not 7a2); the use of h as Correl when it is after Rel.

(ii)                Restrictive favours the use əs- (GTR 1b).


This is of course not to suggest that the distinction is not worth making;  it is undoubtedly there but is poorly described as Restrictive / Non-Restrictive, which label suggests a wrong link with the Determiner System.  There isn’t any.  Since the distinction in question as well as the Determiner contrasts are more perspicuous in English, I shall offer English examples that contradict expectations normally raised by the terms ‘restrictive’ and ‘nonrestrictive’.


(Ex 71) He carries with him a certain notebook that is black.

             (Det [Indefinite Specific] with ‘Restrictive’ Relative)


(Ex 72) Could you give me any clue that will reduce the mystery?

            (Det [Indefinite Nonspecific Nongeneric] with ‘Restrictive’ Relative)


            Ex 43 illustrates ‘Restrictive’ Relative with Det [Definite Specific] ; cf. Ex 6, Ex 67-70 illustrate ‘Non-Restrictive Relatives going with Det [Definite Specific].  They go with other types of Det also.


(Ex 73)             He puts too much trust in human-beings, who are a bad lot any way.

            (Det [Indefinite Nonspecific Generic] with ‘Non-Restrictive’ Relative)


(Ex 74)             Get me some brown paper, with which I shall wrap this parcel  (Det

             [Indefinite Nonspecific Nongeneric] with ‘Non-Restrictive’ Relative)


            A better labeling than Restrictive / Non-Restrictive Relatives will be Inbuilt and Appended Relatives.  Indeed Appending is very close to Paranthesis.  Inbuilt Relatives are a part of the message carried by the sentence that dominates the FP that dominates the Inbuilt Relatives.  S and S1 are linked by the logician’s ‘such that’.  S presupposes S1.  Appended Relatives, on the other hand, offer information or opinion that is incidental to the message.  They cannot, therefore, be looked upon simply as camouflaged and Coordinates.


(GFR 5) Appended Relative

              S0   Appendage S S1

              Where: S and S1 have FP and FP1 respectively such that FP and FP1 are

              Coreferential and have Determiners that concord in respect of

              Specificity and Genericity.

              Logically, as with coordinates linked by and, S0 entails S1.



(GTR 13)             Appended Relativization opt

                Opd: S specified by GFR 5

                Opn: S […FP [S1 [….[j*-FP1] ….] FP ΄] …]        

                                    Where: j* = j [Appended]


            What about Definiteness?  GP1 (i.e. Rel) always carries Det [Definite], while FP ΄ (i.e. Correl) may or may not carry Det [Definite].  This is just the reverse of Inbuilt Relatives, where Rel may be Definite or Indefinite, but Correl has to be Definite.  Inbuilt Relatives can be left as they are.  Some of the Relativization Transformations will have to carry conditions that differentiate between j- with and without [Appended] ;  this should account for the sporadic overt marking.


            Thus, while Specificity and Genericity with their three-way contrast (Specific, Generic, Neither) bear some ‘deep’ semblance to the logical quantifiers (iota, universal, existential), the other two pairs Definite / None definite, and Inbuilt / Appended define the cohesion of communication, the quanta in which the universe of discourse and the truth claims are being staked out.  The first pair is more Semantic, the latter two pairs are more Grammatical.




It’s time we draw a moral


            First, language universals must be hard-won to be genuine.  Methodologically, we must proceed as if we are not very hopeful or even anxious to find language universals.  If we still keep running into some, they must be worth celebrating.  If Marathi or Hindi are not varieties of Sanskrit, nor are they dialects of English with trivially different surface-realizations.  As long as there is no clear evidence to the contrary why postulate [Det N S] for Marathi rather than [S Det N]?  Just because English has [Det N S]?  There must be enough appreciation of the idiosyncrasies of a language.  The close association of the Indefinite k- series with the Relative j- series and the occurrence of unsimplified sequences by the side of those achieved by GTR 7 will then be given their due.  The reward of an honest ‘data-orientation’ is an unexpected clarification and strengthening of our model language.  Thus, while the Marathi Determiner system is pretty murky compared to the English counterpart, the reverse is probably true of the Relativization System (including Appositives – sentential or otherwise).


            A related point is the recent neglect of ‘surface’ possibly, a pernicious result of taking the deep-surface metaphor too seriously.  Listening to some recent argumentation, one would come to believe that language takes some perverse delight in camouflaging ‘deep’ semantic structures by ‘surface’ syntactic structures!  The ‘transformations’ that take us from the depth to the surface must be as few and simple as possible.  Further, they must make some sense in terms of the communicative function.  Consider, with this thing in view, the Antecedent-Copying of Junghare in comparison with the Correl-NP’ Residue in the operatum of GTR 3b; consider also GTR 3a, b and the conditions thereof.  If two forms look different, you must assume that they are differently used till the contrary is proved.


            The Lexicon is much more than a list of morphemes and idioms; it is a list of all the irregularities (as Bloomfield would say) or nipātas  (as ancient Indian grammarians would say).  The GFR and GTR capture the basic generalities, the ‘significant generalizations’ (the sāmānya kāryas) in terms of the Operand, the Operation, and the Context.  The Conditions and the LTR plug the leaks and mop up the nipātas.  Writing grammars is, like politics, the art of the possible-and not the art of sweeping things under the carpet of elegant design.


            In writing a paper rather than a whole grammar it is quite legitimate to carve out a chunk for intensive study.  Unfortunately this self-imposed myopia may lead the analyst to miss seeing the woods for the trees.  If we tried to guess from the expanse of the territory investigated the elaborateness of the resulting map, we have to reverse the architect Gropiu’s dictum and say, More is less.


            But this is not the whole story.  Not only is there probably an optimal size for the chunk.  The methodology of presentation has to observe a somewhat opposite principle.  The description of language should be cast in parts or subparts (or components, to use the currently used hardware metaphor) in such a way that each part and subpart is as self-contained as possible.  In the maintaining of cost accounts, imposing ordering conditions (apply before or after, don’t apply if) over and above the inherent order, imposing elaborate structural conditions on the Operand (especially conditions calling for the ‘or’ brackets { }), or imposing Context sensitivity add up to the cost in that they add to the wiring.  If we can make the Relativization.  Transformation Family independent of the problems of Pronominalization, Word-order, and so forth, it is a net gain.  Thus, if Relativization analysis doesn’t force you to take up a position on the status of Pronominalization or on V, S, O order, don’t.  But, honestly, this is more than the methodological wisdom of not putting all one’s eggs in one basket.  There is also an underlying belief about natural language.


            Language is probably more like a federal structure whose parts are relatively autonomous than like a monolithic structure where “tout se tient” with a vengeance.  Such a federal structure will account for the fact that language can be learned piecemeal.  Hence too the persistent appeal of hypotheses claiming at least a qualified autonomy for phonology (and phonological change), for grammar, for lexicon, and for semantics.  When Harris showed by practice that grammar doesn’t have to worry about phonology all the time as Trager and Smith seemed to claim, or when it was shown that a componential analysis of certain vocabulary domains like kinship could be usefully carried out without worrying about the rest of language, or when syntax was separated as a ‘deep’ matter from ‘surface’ morphology – there was a sense of relief.  On the other hand, the respective downfalls of autonomous phonology and autonomous grammar, the inroads of recursivity into the neat separation of ranks (size-levels), the exposure of the inadequacy of the building-block view of the phoneme – all these have left the linguist a little uneasy.  But he need not feel uneasy, if we look upon these convulsions not as a blow for the monolithic view, but simply as rejected false starts in correctly locating the natural seams of language.  (After all, the division between concrete and abstract phonology is staging a comeback – at a slightly different place.)


            But of course we are not drawing several morals.  All this adds up to a single moral.  The unity of linguistic analysis is the unity of a basic tool kit and not the unity of an excessively constrained and constricting model of whatever persuasion.




            My sincere thanks are due to Mrs. Indira Y. Junghare for providing me with the stimulus and to Mr. V.S. Dongre for bringing –s-GNoC Formations to my notice.


            The recent use of ‘Appositive’ for ‘Appended’ is careless, and probably due to the fact that Appositive Relative Phrases (other than Sentential Appositives) tend to be Appended rather than Inbuilt (as in Ex 68).


            Those interested in evolving a modern Indian Sanskrit-based terminology may consider the following suggestions.


            GFR vigahana-sūtra

            GTR purangaṭhana-sūtra

            LSR  parigaana-sūtra

            LTR nipātana-sūtra

            Operand sthānin

            Operation kārya

            Operatum ādeša (Opd as it appears ‘after’ the Opn)

            Context  nimitta

            Where :  jñtavya

            Condition  šart

            Obligatory Operation  nitya kārya

            Optional operation  vikalpa (-kārya)

            Example udāharaūa (which ‘confirms’ the rule)

            But: / Counterexample pratyudāharana (which ‘escapes’ the rule)

            Exception  apavādabhūta (which ‘disconfirms’ the rule and therefore calls

     for modification in Opd, Opn, Context, or Condition so that it is

     converted into a mere counterexample)

            Definition paribhāa (e.g. Our  definition of Opd 1, of Pc and so on)

            Relative sabandhī

            Correlative sabandhita

            Inbuilt anusyūta (sewn-in)

            Appended anuagika

            Relative phrase (attributive / appositive) sabandhi padabandha (višeavat /


            Relative Clause sabandhī upavākya




Junghare, Indira Y. 1973.  Restrictive relative clauses in Marathi.  IL 34 : 4. 251-62.


[Received 27 December 1973]




This was published in Indian Linguistics 34:4:24-300 December 1973.