Toward a grammar of syntactic change

Lingua ~k~ (1969) 1-27, ,~ North-Holland Publishing Co., Amsterdam Not to be reproduced in any form without written permission from the publ/sher TOWARD A GRAMMAR OF SYNTACTIC CHANGE ELIZABETH CLOSS T R A U G O T T There is an extensive body of literature on the subject of formalized theories of phonolggy, morphology and semantic change, but regrettably little on the theory of syntactic change. An exaggerated picture of the situation may be given by such a textbook as Lehmann's Historical linguistics: an introduction (1962), which discusses genetic change, if only briefly, st every level of language other than the syntactic. All the same, the bias of this book is symptomatic (cf. the lack of syntactic material in Hoenigswald's Language change and linguistic reconstruction (1960)). The present article attempts partially to fill the gap from a transformational point of view and to point to some of '~he "kinds of change found in sentence structure excluding lexicon. Particular emphasis will be laid on the two processes that are usually considered the fundamental universals of linguistic change: simplification and elaboration. No attempt is made a.t all-inclusiveness. Only a limited number of possible types of change are discussed; it is unlikely, however, that other changes would prove to be of fundamentally different character. The observations and hypotheses put forward here can hopefully be tested against tl'e histories ot many different languages so that a universal granunar of syntactic change may eventually be developed. 'rhe examples here are all taken from the history of English. t) The conventional designations Old English (OE), Middle z) The bulk of the prirrmry data is selected from: (i) King All:ted's Orosius (c. 880). Ed. H. Sweet, E E T S No. 79. Lolzdon, Kegsn Paul, 11383; (if) The works o/Geo//rcy Chaucer (latter half fourteenth century). Ed. F. N. Robinson. New York, Houghton Mifflin, 1957, 2.~d ed.; (iii) The complete worhs o/ Shakespeare (end of sixteenth, beginning of seventeenth centm-ies). Ed. H. Craig. Chicago, Sco~.. Foresman, 1951. Other quotations are selected from the following worxs' ~v) The romance o/ Sir Beues o/ Hamtoun (c. 1327). Ed. 2 E L I Z A B E T H CLOSS TRAUGOTT English (ME), Early Modern English (ENE)and Modern English (NE) are used as cove: terms for the periods 500-I 150, 1150-1500, 15001700, 1700-present, re:.~pective!y. It should be noted, however, that the traditional sharp distinction between ME and ENE is one made on the basis of phonolc,~,cal, morpholof,~cal, and orthographic changes; from a purely ~yntactic point of view, ME after 1300 and ENE are remarkably similar; the major changes syntactically can be located in the OE, early ME and NE periods, not in ENE. HISTORICAL CHANGE AS REF.LECTED IN SYNCHRONICGRAMMAR Before turning to the main investigation, it might be useful to summarize vh:~vs on a question that has been much discussed recent.~y: to what extent do diachronic and synchro,aic grammars overlap ? A synchronic grammar is in no way a model of a static, monolithic chunk of language; rather, it must of necessity be 'composed of chronological layers' (Kurylowicz, 1962: 9), some more archaic, some more innovating. This is the inevitable result of language acquisition, dialect interference and the many other external as we:] as internal features that affect individual languages. It is just this multiple layering that has often been used as a starting point for internal reconstruction and comparative linguistics in the area of phonology and morphology: 'if we cannot ch'aw diachronic implications Iront synchron;.c structure, we cannot do comparative linguistics' (Teeter 1962: 1031). Well-known examples of such diachronic implications found in synchronic stnwture are raorphophonemic consonant alternations like those in iwuse-houses, breath-breathe, which, it can be shown, are relics of OE when there was phonemically only E. KOlbing, .frETS Extra Ser~Js No. 46. London, Kegan Paul, 1885; (v) The New Testament in English by John Wycli]/e (.z. 1380). Ed. J. For~Aaalland F. Madden. Oxford, Clarendon Press, 1879 (repr.); (vi) Cursor Mundi (c, 1400). Ed. R. Morris. E E T S ~'o. 57. London, Kegan Paul, 1874; (vii) The Paston Letters x422-z5o9 . Ed. J. Gairdner. London, Chatto, 1904; (viii) Le Morro Darthur by Syr Thomas Malo,y (e. 1469). Ed. H. O. Sommer. London, Nutt, 1889; (ix) Pierce Peni!,'sse his supplication to the diuell (1592), in The works o/ Thomas Nashe, Vol. I. Ed. R. B. lffeKerrow. Oxford, Blackwell, 1958; (x) The Holy Bible, ~uthorized version (1611). Okford, University Press; (xi) The li]e and correspondence o] the late Robert Southey (end of eighteenth, beginning of nineteenth ceuturies). Ed. C. C. Southey. London, Longmans, 1849. TOWARD A GRAMMAR OF SYNTACTIC CHANGE 3 one spirant series. Voiced and voiceless were predictable roughly as follows: voiced medially, voiceless elsewhere. Various criteria, most importantly simpl~icity, suggest that morphophonemic rules accounting for such consonant alternations as house-houses, or such vowel alternations as serene-,.:emnity should be ordered in a synchronic study of modern Englisil (of. especially Halle 1961, 1962). This order has until recently been established entirely on metatheoretic grounds. The ordering is strictly logical, atemporal, But it has also been noticed that often such ordc,'ing actually reflects historical change. This observation is not new; it can be found in slightly different form in the following much-quoted passage from Bloomfield's 'Mt;nomini rnorphophonemics' (1939" 106)" ' T h e p r o c e s s of d e s c r i p t i o n l e a d s us t o s e t u p e a c h m o r p h o l o g i c a l e l e m e n t . . . . . . Lue u e v l a ~ o u ~ f r o m u , ~ b a s i c in ~t uleor-eu~al b a s i c l l . ~ I l l l , ~IIU t h e n t o nt~Le f o r m w h i c h a p p e a r w h e n t h e e l e m e n t is c o m b i n e d w i t h o t h e r e l e m e n t s . . . O u r b a s i c f o r m s a r e n o t t h e a n c i e n t f o r m s , s a y of P r o t o - A l g o n q u i a n . . . h o w e v e r , o u r b a s i c f o r m s d o b e a r s o m e r e s e m b l a n c e t o t h o s e w h i c h w o u l d be s e t u p f o r a d e s c r i p t i o n of P r o t o - A l g o n q u i a n . . . s o m e of o u r s t v . t e m e n t s of alternation ... resemble ... as to content and order, the histor,cal development from Proto-Algonquian to the present day.' This concept has :ecently been elaborated and made ~ matter of great theoretical cor,cern (e.g. Halle 1962; Stockwell 1964; Saporta 1965; Kiparsky 1967, 1968; Goman 1967). It has been sub.,gested that where two orders are both possible, the one reflecting historical order should be selected. This is not to say that synchronic and historical order are or should be the same (for explicit rejection of such a hypothesis, ct. Chomsky 1967 : 127 ; Chomsky and Halle 1968: 251; Kiparsky 1967:91); they may, however, coincide, and if they do, considerable insight into language acquisition (Halle 1962) and dialect relationships (Klima 1964b) may emerge. Furthenmore, they may give a measme of psychological reality to the principle of ordered rules in synchronic grammars (Kiparsky 1968: 181). It is not only ordering a~,~a metatheoretic principle that may be given psychological reality by historical evidence; .,,o may other constructs, such as deep structure relationships, features, etc., established to account for sy~.chronic phenomena, but not necessarily having any overt expression ~tt a given period. Katz and Postal (1964: 97) note that their analysis of 'yes-no' interrogatives as wh- 4 ELIZABETH CLOSS TRAUGOTT questions of the structure wh-dt~er or S + S has reality not only in the 'indirect question' form I asked him wlw,ther l~ c~t~ vs. the 'direct question' form Did he come ? but also in the opti_cr~alOE yes-no question introducer hwa~er (wh + aegcter : wh -t- eitb.er]~ o/ ~ ) , cf.: Alfred, 0r.220.8: Hwae/~er Romane hit witen nu aenegum men to seiganne hwaet biers tolces . . . forwurde? 'Would the R o m a n s now blame anyone for saying how many of their men . . . fell ?' with hwaMer, bes:iLde: Alfred, 0r,94.30: 'Ne gol~ync6 ~e swelc gewin noht lustbaere ? 'Does such conflict seem to you in no w a y desirable ?' without. Again, we fin,! in OE ample overt support for the hypothesis that certain verbs are inhermtly negative. In NE indefinite quanfitiers like some, someone, somebody and indefinite 'time adverbs' like sometimes supplete to any, anyone, anybody, ever respectively under specifiable conditions, including constituency of complements in negative sentences, e.g. He doesn't a//ow me to see anyone. Suppletion of some by any etc., also occurs in the complements of verbs like doubt, re/use,/orbid, d~sli~e, etc. even though neither the matrix nor the complement sentence contain sE~ (sentence negative), cf. ! allo~.,ed you to do something, but I forbade you Lo do any~i#g: He agreed to see somebo~v, but He re/used to see am2body. Since these verbs cause suppletma they may be considered to have an inherent feature [-/- NEG] w.hich triggers suppletion but does not introduce overt neg~.~.ive elemeW,s such as not (cf. Klima 1964a; Fillmore 1967; UCLA 1967: Negation). In OE such verbs, however, not only required the suppletive forms of the quantiiiers in object compl~.ments but also optionally introduced art overt negati re in a semantically affirmative subordinate sentence :: Alfred, 0r.262,22: Fo~bead ~aet mort na 6aer eft n¢ t::.~brede. ' F o r b a d e t h a t one hover there after not built --= Forbade anyone ever co build there after.' with negative format~ve~,, in the complement, beside: Alfred, 0r.254.8: F o r b ~ / | ) a c t hiene mort god hete. ' F o r b a d e t h a t him one called God ---- F o r b a d e t h a t a u y c n e should call him God.' TOWARD A GRAMMAR OF SYNTACTIC CH.~,NGE without. Both of the changes, which led to the loss of overt markers of deep structures, illustrate a general tendency in v.he history of English for deep elements to become more and more covert as tar as phonetic realization is concerned. The discussion in tl'As section has so far been devoted only to the kinds of supplementary information and support that diachronic grammars are well known to provide for synchronic grammars. As historical evidence is brought more and more to bear on theory of grammar, particularly that concerned with universals, it is likely to play an increasingly important role, even if it is always a seconda~r one. Few recent Sl~culative works ignore di~chronics completely. For example, when he suggests that P(hrase) S(tructure) rules should be unordered, McCawley (I 968) brings supporting evidence not onl,, from mathematical models and synchronic grammaTs but also from indirect historical evidence: e.g. the fact that no examples have a~ai,.a~ ~ l ~ a ~ - ~ , L ~ a ~ a ~ a a ~ . ~ o L~o~a~ga$ from ~UA~L~tlL,~' . . . . . . 1. 1¢4[ aXll--o r d e r - ing between PS rules wherees many such differences have been found that result from varying orders of T(ransformation)-rules. If no such dialect differences occur, it is at least partly because no reorderi,lg of Ps rules seems to have occurred historically. If historical e~idence supports the theory that the PS need not, indeed should not, be regarded as an ordered set of rules, it also supports the theory that PS rules need not be internally ordered trom left to right. In ray work on the his~.ory of English, I have constantly been p;~agued by the problem of word order. In oE the usual order is [HP [AUKV NP]vP~8; in co-ordinate and subordinale clauses, however, it is [sP [NP v AUX]vr]s. While departures _~rom these two o~'ders frequently occur, most of them can be shown to indicate 'raarked' structure (cf. especially Bacquet 1964). In terms of a synchronic grammar of oE, it is simpler to do what Bach (1964) suggested for Modern German and give the subordinate order as the basic one. A T-rule would convert this subordinate order into main clause order in appropriate structures. We could then postulate that this order-switch rule became dominant in ME unt:il it eventually led to a change in Ps order, such that by ESE we have only Ear ~ u x v Nr]Vr]S order. The trouble is that such a change would seem to be ~,e~ ~ignificant in terms ~ff the gramma~ (i.e., result in extensive radical mutations) ~s it involves change~ in the deep structure; however, no change in the iunction of word order accompanies 6 ELIZABETH CLOSS TRAUGOTT such a putatively deep alteration of structure• In both OE arm ME, order is the prime signal of function; even if the OE case system allowed for relatively more freedom than was possible in ME, the functional load of order was very high, as Bacquet convincingly shows, there is no evidence to support the hypothesis that oE word ordex is incidental and merely a function of co-occurrence restriction, whereas ME order is functional (as well as having incidental cooccurrence restricti.ons). B at only if such a hypothesis were proved true could we justify claiming that there was a deep structure change. One possibility would be to say that OE and ME word order were the same in the underlying structures, and to specify orderswitches for co-ordinate and subordinate clauses in the T-rules. The only difference between oz and NE as far as ord,:r is concernc, d, would then be that the order-switch rule is deleted by ENE times (it persisted throughout ME, bat became increasingiy -~,~o;,'ox Another possibility wcul"., be to reject the metatheoretic principle that constituents within PS rules should be ordered from left to right. A PS .~ule would then generate unordered categories and relation,;hips, i.e. [NP, VP]s would signify both [NP--VP]s and [VP--NP]s, and [Aux, v, NP]Vl, woud signify all logically possible orders available to the three coastituents. Ordering (including exclusion of impossible orders) world be effected by the T-rules only. Similar su~ges,tioas are currently b,~.ing discussed with particular reference to ,deep ,'ase (cf. espe~ ;,ally' Fillmore 1966); the theoretical arguments Filhnore provides may be given added dimension by diachronic evidence of the relative lack of innovation in deep case relationships (implying deep structure membership) as opposed to relative frequency of order changes (implying surface structure membership). TYP,'x,,BOF SYNrACTIC CHANGE Like those in other areas ef the grammar, most syntactic changes can bc grouped under the two opposi.ug tendencies of simplification and elaboration. To separate these t~o completely is difficult since they are often involved in the development of one and the same strucl~:ure. Furthermore, simplification in one area may trigyer off elaboration ,'lsewhere. On the other hand, unless the tendenc~'es are at least initially separated, exact processes of change cam,ot be adequately ~.ccounted for, nor can the substantive differences and TOWARD A GRAMMAR OF SYNTACTIC CHANGE 7 • imilarities between phonological and morphological as opposed to .,yntactic change. In his discussion of phonole~:ical change, Kiparsky (I 968: 200-202) has shown that simplificatio~t is not of one but at least two different kinds, depending on which F~rt of the rule it affects, i.e. depending on whether it affects the structure index (that part of the rule specifying the elements to which, *he rule is applied) or the structure change (that part of the rule s[, :cifying what happens to the elements given in the structure index ~. He points out that simplification of the structure index leaves no r :lic forms in the surface structure and therefore is not recoverable from synchronic grammars; furthermore it results in what is traditionally known as analogical extension or l,olarization, where a :~ativel ~ rare form is used in more and more contexts. On the other hand, aiteration in the structure change may hiive relics in synchronic gramr lar and may therefore be recoverable; it results in what is traditicnalli¢ known as analogical leveling, wh.~re a form is used in fewer and fe~,~r contexts, sometimes being eventl~ally lost altogether. It is usefifl to dis, inguish c:hanges in thc structure index and structure change of the syntactic component also, although the different status of the elements under discussion (gramn.~cal elements vs. pl:onological feature matrices) inevitably aeads to some differences in the type~ of change involved anti their effect. The terms structure index and structure change will hereafter be symbolized as sl and sc respectively to avoid possible confusion between historical change and derivational, atemporal 'structure change'. The word 'change' will be used only to refer to historical processes. Changes in sI and sc in the morphophonemic component provide an analogue ,.rely to the transformational component of the syntax since they ar~: concerned with changes in the operations applied to strings; they do not provide an analogue to the deep structure component of the syntax. Since the PS specifies constituency and relationships, changes in this component involve not changes in m~mipulation, but alterations of the underlying structure of the grammar itst~lf Ideally, the es of some grammar may be regarded as ~. subset of a language-universal set of categories and relationships (ct. Katz and Postal 1964; Chomsky 1965); hence changes in the I,s ideally involve the development of a new subset of tlhe universal grid. From the point of view of formal universals, T-rules can also 8 ELIZABETH CLOSS T R A U G O T T be Shown 'toform part of a univer~d set insofar as they are classified according to type (left-sister~dju~ction, deletion, etc.) or a¢'cording to operation (F-relative-attachment, etc.). O n the other hand, the actual de,l:ailsof T-rules (e.g. which element is some constituent attached ,to as the left-sister?) are for the most part language- specific, so whellt we speak of changes in T-rules we will often be referring not to changes in the subset of universal rules but rathcr to changes in the particular specifications in the sl or sc. Therefore, as far as the grammar as a whole is concerned, changes in the PS ~xe on several different counts functionally dJifferent from changes in the T-rules. Simplification and elaboration can all the same be s,.id to be the main kinds of historical processes affecting the total grammar; the result of these ch.~mges will depend on which component they apply to, but the type of change remains constant. Within the history of English there are few instances of PS simldification, if we assume that the PS involves rules such as those exemplified in Chomsky's Aspects o ~ the theory o.t syntax (1965: see especially p. I02); perhaps fewer and fewer changes will emerge as the PS comes to be further and further refined with grammatical features including [TENSE] (cf. Ro~enbaum and Loc~aak 1966: 9), logical relationships (cf. Lakoff 1965, 1966; Bowe~ ! 967), gr~mmatico-semantic features (cf. Klima 1964a; Weinreich 1964), ~tc. Be this as it may, some reductions of the PS can clearly be traced. One is the loss by the end of the ~E period of the ir,~initive marker .INF required by all modals; but as this change affects only the phonological shape of verbs following modals, not the relationship between constituents, nor their function, INF should probably not be given deep structure status. More interesting are changes in cert~-lin fe~,ture-hiera:cchization:-~ and feature assignments to categories. A well-known .instance ~. provided by the history of gender distinctions in Englis.i*L OE wr~scharacterb'ed by grammatical gender, at least as far a.,; the nomiJ~,al ~ystem was concerned. A noun in OE must be marked, among other things, for: (a) ,! "4" ANIMATE t {+ GE D Z ] (b) + GENDER ± CC) -MASC T O W A R D A G R A M M A R OF S Y N T A C T I C C H A N G E 9 etc. where [ + GENDER] and [ + ANIMATE]are simultaneous features and where [-- MASC] is 'neuter'. Personal (and also relative) Proforms of these Ns follow either this system or the following alternate system: (i) :~: ANIMATE (ii) 4(iii) -b (iv) ANIMATE ~ + GENDER GENDER --~ -~ FEM -- rEr~ ~ + MASC (V) --" ANIMATE - - ~ - MASC As ilh:stration, in r~. the animate N wi/'woman, wife' was C-- MASC~ by r,.tle (c); determiners aad adiectives agreeing with w i / h a d to be marked for [-- MASC]; but the Pro-form could be either [-- MASC,, realized as .bit 'it' or, mole often, [ + FEM] by rule (iii), realized as bed 'she'. Conversely, the inanimate N burh 'city' was [ + FEM~ by rule (b), but the Pro-form could be ~-- MAsc] by r ~ e (v). During MEthe set of features assigning syntactic gender was greatly reduced; by ENE the gender system came to apply to nouns only insofar as pronominalization rules were concerned and it lost all relevance for determiners and adjectives or for nominal case endings. The only relic of morphological gender markers on Ns that remains is in some 'feminine' derivative suffixes like -ess, -ette. The Pro-N gender system is itself essentially that of rules (i-v), so we now have a system of 'sex gender' not 'grammatical gender', i.e. a system applicable only to Pro-forms of animate Ns (barring a few relics of the grammatical gender system which still persist for ships, cars, etc., the usage often depending on the sex and professional interests of the speaker). I*~ appears tb,m that for OE every N must be marked in two ways, one for the gender system applicable obligatorily to its nominal and optionally to its Pro-form, the other ft,r the alternate gender system applicable only to its Pro-form; by ENE the first gender system is lost and the second becomes obligatory. Another example of changes in feature assignment to categories can be found in the specification of [± DU;~r~], but this time the char~ge is tess far-reaching. [ + DUAL]is still available for quantifiers (cf. both), for disjunction (cf. either-or), for prepositions (cf. baween), but no longer for Pro-Ns. In OE there was an opposition between ic 'I', wig 'we two' (dual as opposed to we twegen 'we two', which forms part of the series 'we three', 'we four', etc.) and we 'we' (more than 10 E L I Z A B E T H CLOSS TRAUGOTT two, or unmarked as to duality); there was also an opposition between ,bu 'tl~ou', 5it 'you two', and 5e 'you'; no dual form of the third person Pro-Ns e~:isted, however. Duality as a fe?.tu~e of the personal Pro-Ns dropped out completely b y ME times; unlike the dropping out of tla.e nominal gender assignment, this ctid not affect the grammar as a whole, only a very limited part of it. Another kind of es simplification evidenced in tile history ff Engli.,~h concerns generalizations from mutual exclusivity to mutual co-occurrence. In oE there are no data to suggest that modals an,i perfect aspect could co-occur. In other words, such a sentence as: *He wolde (MODAL) ge/eohten (MV) ~abban (PERFECT). 'He would haw. fought.' is apparently unframmatical, whereas there are hundreds, indeed thousands, of constructions like: Alfred, Or.8~;.28: Ic haabbe (PERr~Cl) nu gesaed (My) hiora ingewinn. ' i have nov., sc;id/told of their ancient battles,' Alfred, 0v.21.4: And hyt motan (MOr.AL) habban (~v) eall. 'Anti ,r~st/may keep it all.' Furthermore, progressives appear t a co-occur only with mod~!~, not with perfects: Alfred, Of.274.19: l~a ,,;ceoldon (MOD.',L) on siml .~.aon (PROGRESSIVE) u ~ nende (MY). 'Thes~, should/had t,, alw~.y, ,~e fighting.' *loa h~fdon (P:i~RFEC;I) on siml g,et:.Je, e n (PR,)GRESSIVE) tvin#~'nd$ (MV). 'These had Mways been fighting. This suggests tllat, in terms of ruL~,s such as are used in Aspects the OE Auxiliary should be specified ~s: AUK --~ TENSE y(MODAL) (PROGRESSIVE) 1 [(PERFECT) I where MOD.~L is a cover term for M - INV. i.e., for the formatives cunn- 'can' mag- 'may', mot- 'must', soul- 'shall', will.. 'will', each requiring the infinitive marker on the following verb; and where PROGRESSIVE is a cover term for beo. be', w,~'s- 'be', weerd- 'be', each requiring the present participl~ mar:~:er on the following verb. By ~tE however, we find MODAl. followed by PvrFECT, as in: Chaucer, T. of Mel. 2287: David the king, th,Lt wolde (MODAL) have (PErFEET) slayn (~V) him. 'David the King, who would have slain him' (or is wolde the main verb will- 'int,~n~' ?). TOWARD A GRAMMAR CF SYNTACTIC CHANGE and rarely, PERFECT f o l l o w e d ]I b y PROGRESSIVE" Chaucer. K T.929: V(e b*m (P~.~F~'CT) ben {'PROGttESStVZ) wa~tynge (MY) al this iourt~mygkt. *We ha~e been waiting all this fortnight:, ' It appears, then, that the M~ r~le for Auxiliary was very much like that of ~ . Besides other eletnems, to be discussed below, we presumably have the ~ u e n c e : AUx-, TENSE (MODAL)(PE F :CT)'ROGRF.SSZVE) ( for ME. In other grammars where m~]alities and aspects are generated either as sentence Mements (possibly part of presentence) or as main verbs (cf, McCawley 1967; Ross, 1967; Boyd and Th orne 1968; Traugott and Waterhouse 1968) these differences may be expressed by restrictioas on hierarchizalion of sentence elements or by constraints oli embedding (if the lat~er, tee changes affect not the PS but the sO. Regardless of the particular analysis, the historical changes involve generalization and hence simplification. The development of PERI~ECT is itself interesting, though descriptive details again depend on the particular form of the grammar used. If Imve - P ( a s t ) P ( a r t i c i p l e ) in NE is considered the re,~lization of abstract deep PERFECT~then the changes ~scussed below ~re part of the lexicon and morphophonemics and should not concern us here. But as long as grammars introduce grammatical morphemes into the ps, changes involving them must be considered part o~ the deep structure. In OE, PERFECTin the environment of transitive verbs was habb- - vP. I~i the environment of intransitives it was usually one of three verbs ~vhich required ~,P; these are beo- 'be', w e s - 'be', and weordo 'be'. A few intransitiw2s like fitr- 'go' also allowed habb. - PP. During the ~istory ot English, the latter became generalized to most verbs, whether transitive ot not, although a great number of intransitives cmld optionally select 3e - PP as PERFECT a,; late as the seventeenth century~ In fact some, notably come, still select be - PP almost exchlsively in Shakesi)earean drama and in the Authorized version of the Bible, cf. Shakespear,~: Merry Wives III, .49: We are c~,ne to you to do a good office. Bible: Isa. 10.28: He is come to Aiath, he is passed to Migron. beside the rare: Shakespeare: Ant. and Cleo, II ~ii. ! 1 : V¢ould I had never come from thence. ]2 E L I Z A B E T H I CLOSS T R A U G O T T Bible: Isa. 7.17: The Lord shajL[ bring upon thee . . . da,,s tb.at have not In both Shakespeare and the ]3ible have is used with come and similar intransitives almost exclusively in negative sentences. What the significance of this constrairt may be is as yet unclear. In NE we have be - PP (often obligatolily) only with adjectivalized 'ststives' (of. b e / r o z e n ) and verbs with inherently built-in perfectives (cf. be born), both of which may bq~ historically related to the perfective be - PP, but neither of which is a direct reflex of the regular PERFECT as discussed here. The two types of P]~RFECT,which in DE had almost equal functional load, moved in opposite directions, both involving simplification; in respect to l~abb- - PP this involves generalizatiol~, whereas it involves various stages of loss in respect to the other formatives (weor~- - PP wao totally lost by the middle of the ~ E period; beo- - PP and w e s - - PP both remained in some dialects including Shakespeare's [cf. they be vs. they are, both as indicatives] until lthe end of the ENE period and then, for Standard English at least, became one suppletive verb be - PP which in turn was lost as the regular realization of PEr~FECTin the environment of intransititires). If we turn from simplific~ttion of the Ps to simplification of the s! we find that it too always involves some kind of generalization, or else total loss of some restriction or form specification. Generalization can from a purely surface point of view be considel"ed elaboration or addition, since it involves extension to wider and wider context~; on the other hand, from the point of view of the internal grammar itself, generalization is simplification as it involves reduction in the number of restrictions. Further details from the auxiliary system may suffice to illustrate sI extension. In oE, in addition to TE~SE, only modals could occur in the sl of passives, but not perfects or progressives. Thus we find: Alfred, Or. 128.5: ])a Darius ~ ~,seah ])set he o/erumnz~,. (MV)b$Oiq (PASSIVE I a MARKER) wolde (MODALI~ V~hen Darius saw theft he would b8 do. leafed." but apparently not: * pa Darius geseah ])set he o/evwunnen (MY) gsweset~ (PASSIVE MARKER) h~f~ (1°ERFECT).'When Darius saw t h a t he had been ¢le/sa~d.' * ])a Darius geseah ])eat he o/erw~nz~# (MV) W~S¢~'. (~.SSIVE MARKER) waes (P~:OO~ESStVE). ' W h m Darius saw t h a t he was ~ / z ~ d~/ea~d." T O W A R D A GRAMMAR OF S Y N T A ~ T I C CHANGE 13 By ME perfects t~came available too: Chaucer, T. o] M d . 2210. By cause of the wrong and of the w:kkednesse that h~A (PSRWECT)5 s ~ (PASSZVEMARKER)d00N (MV). ~HeC~US@of the wrong and wickedness that has been done.' Progressives finally became available too by the end of the eighteenth century: So~, Vol. 1,249.24 (Letter 9th Oct. 1795i):Like a fellow whose uttermost Upl~)r grinder is being torn out by the rc~ots by a mutton-fisted barber.i) The chan~es are reflected in such sfs (for "r-pa~ive) as: (a) for o~:: x NP TENSE (MODAL) V NP PASSIVE Y I~) 3 45" 7 (b) for X NP TENSE (MODAL) (PERFECT) V NP PASSIVE V ME andENE: I 2 (C) for NET 3 45 6 7 X NP AUX V NP PASSIVE Y 123 456 7 where only the constituents of element 3 change, an,t where the sc remains constant except for a few details concerning the grammatical morphemes "vhich function as passive auxiliarie:~ and agentive prepositions (tt, be discussed below). It is particul;Lrly important to note that while these changes match phonological sl changes insofar as they are non-recoverable, sI extension in the transformational component involves not so much t,xeater use of a rare item, but rather extension to cover higher l~odes in the PS. Simplification of the sc may ir.volve reduction of additions if the T-rule introduces adjunction or tt may involve more extensive de}etion if it is of the deleting type, etc. One seemingly obvious kind of •;implification is that involved in the reduction of grammatical formatives introduced by The sc's. It was mentioned above that passive sc's changed slightly during the history of English. The de~:.ails are as follows: in os there "~vere three passive auxiliary forma. ~:ives, beo-'be', wes- be', weorar- 'be' homonymous with the mtransis) 0 u o t e d by Moss6 (1938: V0d. H. 48) as the first authentic example o~' '~:his construction. ]4 ELIZABETH CLOSS T R A U G O T T tire perfects and, like them, requiring PP; several prepositions could function as agentives, notably/tom, ~i~, l~urh, though malay others were also available (cf. Dam 1957: 23). By ME the passive auxiliaries, along with those (unctioning as intrarsitive perfect and progressive auxiliaries, were reduced to two by lo:~s of weord-, and the agentives were replaced by 7mith and by (both stil 1 used at Shakespeare's time). Not until NE did the single formatiw:s be -- PP and by Lecome established. In grammars where abstract element:~ like PASSIVE AUX and AGENTIVE are introduced by adjunction transformations, such changes are strictly speaking not a m ~tter of sc, but of lexical and morphophonemic change. The predominantly lexical nature of the changes can be seen clearly when it is noticed that with in ME replaced mid in all con_texts, not just passives, but also in adverbial phrases of manner, accompaniment, etc. In OE there was a preposition wi/~ but it meant 'against'; the reasons for the shift from wi/~ 'against' to u,ith 'by, with' c.,n still be traced in the ambi~Mty of usage in such expressions as 'to light with ---- against' vs. 'to fight with = together with, on the same side as'. Most of the examples discussed so far have been relatively simple in that they involve change affecting just one kind of structure. Often, however, two or more levels of structure may be involved in a series of changes, or relationships between rules fo.rmiug a single set may be altered. An excellent example of complex change is provided by the history o~ surface case assignment in Engli~. s) Nouns in oE were marked not only for gender, but also for number and case. While gender and number are inherent features of N that can be introduced in the PS, case i:, not inherent, nor is it of one type ordy. First, there is [ + NOMIrCATIVE2 which is assignable automatically to the surface structure subject, whether N or Pro-N. It is a function exclusively of the relation 'subject-of' and can be intrc>. duced by a T-rule ordered after T-passive. Predicate nominals and adjectives are also assigned [ + NOMINaTrVE], but ilere the case, is a fuaction not only of the relationship 'predicate complement' but s) I t is :importan~~. to note t h a t the present di~tcussioa com:erns surface 'accidence' only. 'Inherent' or 'internal' case of deep structure :relationships (cf. especi~dly Fillmore~ 1966; Anderson, lC)68) was not ~fffected directly; changes in the devices for realizing this deep case ~.re particularly intere,~ting from the point of view of the interrelationsh ip of case and prel:~sitions, b u t details m u s t await fuxther study. TOWARD 8, GRAMMAR OF S Y N T A C T I C CHANGE 15 also of the lexical verb. All verbs with complements (and prepositions) can be specified in the lexicon as inherently requiring certain cases on their complements. Link verbs and copulas will be marked in the lexicon for [ + NOMINAtiVE] complement; transitives will be marked for [ + ACCUSATIVE](e.g. S¢O-'see'), [ + GENITIVE] (e.g. ~uc'enjoy') or [+ DATIVE] (e.g. hedp- 'help'), where [+ ACCUSATIVE], r + GENITIVF],[ + D.tTlVE] (probably to be hierarchized in a aetailed grammar) are sub-features of F--NCOMINATIVE]. The latter is a function of the relati.-nship 'object-of' and of the lexical item. The case features are mapped onto the complements by an adjunction translo~nation. A third 'ype of case-assignment is needed to account for the f liet that determ:ners and adjectives in OE agree in number, gender and ca~ with no;~linals of which they are sisters (whether by PS rules or by T-rules .~uch as T-relativization). A very low level adjunction T-rule will ,'-nap these agreements onto the relevant constituents from ~he nominals. All three types of case-assignment underwent ch astic simplification during ME, starting with the loss in the lexicon of the sub-features [ + ACCUSATIVE],~+ GENITIVE], [ + DATIVE]. Space does not permit details to be given here of the stel>-by-step chan~Tes in lexical and sc's. Suffice it to say that by E~E only Pro-Ns :ontinued to show overt distinctions between [ + NOMINATIVE] a |d [ - - NOMINATIVE]. The agreement rule was lost, arid the T-rule t h t mapped cases from verbs (and prepositions) or~to complements was amalgamated with the first rule that assigned case as a functional relation only. this rule was generalized to cover b¢,th surface subjects and complements, wi~h the result that ~-- NOMINATIVE] needed no longer to be marked in the lexicon as g.n inherent feature of verbs. The grammar books have for the most part k~pt this rule intact, but elsewhere it was soon modified at least insofar as the interrogative and relative who-whom distinction wa~ lost. In aguE, for example, such exchanges as: Shakespeare, Oth. I ii.,5,3: lago, lie's married. Cassio. To wko ? Shakespeare, CoJ,. I [,i.7: Men. Pray you, who does the wolf love ? Si¢. The lamb. M¢,n, Ay, to devour him, are as acceptable a~; those with whom. So-called Standard English which re~:ains a who-wkw,m contrast is therefore (in this respect) more archaic ,han man 3 dialects of ENE. It is important to note in this 16 ELIZABETH CLOSS TRAUGOTT connection that synchronic: rules accounting for the different dialect usage of Pro-N case reflect historical ordering only if Standard English is taken as the no1 m (as it is in Klima 1964b). Elsev, here in the grammar, as in the part accounting for the use 6f do (discas~sed below), we will find that Standard English is more innovating: and rules accounting f,~r ~ffcrences in usage in other dialects will r~fl~t not innovation, but rather, more archaic stages. Loss of whole rules i~ clearly a simplification of the grammar that does not fall under the heading of either sx or sc changes. A!th ~u~h cha~ages in parts of rules may affect whole sets of structure, they fl:equently do not do so (ct. changes in the passive), and frequendy do not have very not':ceable effects on the surface grammar. Loss of whole rules, howeve:-~ usuaUy results in at least malked surtace differences (cf. loss of some of the gender and case rule ~). Another kind of sinLplificttion that concerns whole g r m p s of rules is reordering. This brings about Lweater utilization of ru~es. Kiparsky finds that a tendency toward such increased utilization is probably a universal in phonology (1968:196). It ~s possibly also one in syntax althol=gh data limited to Engfish syntax is naturally insufficient to be diagnostic. Inustrations of this kind of change ca'~ be found in the area of group gelfitive constructions and cleft s~nteaces. One way of accounting fo: ENE avd ~E construct~o~ fike 'The wile of Bath's tale', 'The Mayor of Iondon's mace' i; to specify grouping and then genitive-attachment. For the well-known ,~tEconstruction: Chaucer, Troilus and Criseyde 2: T h a t was the kyng Priamus (geuitive) sone o[ Troye. ' T h a t was King Priam o/ Troy's son'. Chaucer. Clerk's T: F o l w e t h the Prolog of the Clerkes (ge~ ~tive) Tale o[ Oxenford. 'Here follows the l:'vologue of the Clerk o/G'x: ,rd's Tale'. it seems tha'~ we must require genitive-attachment be:fore grouping. The ENE and NE order allows for more and more gener Jized utilization of genitive-attachment, now [ ~rmi~ting not only [[DET of N]'s N] but also, ~olloquially at least, such sentences as The g,'rl who lives m;xt door's t,oy/riend. Whatever the exact derivatio~ , :. cleft sen° tences, it is also apparent that ME sentences like: Malory, Merle Darthur 134.16: It ben flxe da~loyseles . . . t h a t :~o ~.ame me. 'It is (lit,~rally, it are) th.~ damsels w lo call me t h i s ' Wycliffe, Mt X.20: I t ben n a t ~e chat spekea. ' I t is (literally, it a~el n o t you t h a t speak.' 4) 4) ~ o t h o~ these sentences are qucd:ed b y Vis~er (lft63: Vol. I. 49). TOWARD A GRAMMAR OF SYNTACTIC CHANGE 17 undergo some different kind of verb-agreerner~t ordering than do EYE and YE senten,:es '~f the type I t is the girls who call me this. For thes~, ~.; sentences at ieast we must say that verb-ag,-r,~ement preeede~ it-placement (where it m a y be some ,dummy form replacing the enes/tt~e?,, in a structure underlying T h e dae~sels ave the ones that so n~m~ me), while ~or later twriods verb-agreement must follow itplae~ ment, so generalizing verb-agn'eement to surface structure subjects A d d ~ i o n s ,and dab, s.rations Tl,.eoret ically, ad:titions to the PS would iaclude additions of some p r o p : t r y tssociatrd with the universal grid. whethe: category, f e a t t r e , ol relatiol~ship, ActuaUy, in the hi,,,tory of English no :~uch ,m to have occurred. If the Ps is aUowed to contain lanchan ~es s,~ .~ gatag "-spe~:ific ~o'matives, however, as is the case with most recent AUX -ontains tw:~ more ftrmatiw~s (do and gin, both requiring INF) t h a n does ~he oIi ~ v x ; these function quite differently i r o n a n y t h i n g found in o r . They first appear in poetic texts, do - INF not until the t h i r t ~ , n t h century. They became common in I,ondo~ prose only c. 140). At this period do appears to be mutually exclusive with progr~siw,s, but it can occur with modals and perfects. Elk'g~rd argues t h a t all instances of do - ~.~F after MODAL are causative in Southern texts, but m a n y instances even in the South are at least ambi~u,u~, as to whether they are causative or noc. Certainly do i:,,v i~; common after have - PP, sometimes alternating with do - PP, i,,:~, s:~metimes requiring a p~st participle marker instead of INF. S~nc, n.~ difference in function seems to be involved, anti since PP n.q,:~ired by do can be considered a r e d u n d a n t participial marker t~igg,~red as a surface phenon~enon by the PP of PERFECT. do - PP is not con:dd,~'red as PS element ,;eparate from do INF. Rather, the PV will be optionally introduced by a T-rule accounting for various redundan,:ie,=; (~uch as double :,uperlative, etc.). Altho~.gh l have not been able :o find unambiguous examples of gin - ~NF a:ter NODAL or PERF~;C~;~, or in code, it nevertheless seems~ justifiable ~:o generate it in the same way as do . i s F i n a t u l e f o r ~ E l i k e : - .'[8 ELIZAEETH since same do oo Some CLOSS TRAUGOTT it frequently alternates with do - INF in different Mss of the text (for many instances of alternate texts with do oo gin, or gin oo no a'tudliary or gin c~ no auxiliary, see Funke, 1922). examples of the t ~¢o auxiliaries are: Chaucer, Sv, mmo~er's T. 2042: And thus he dide doon sleen hem alle thre. 'And thus he caused them all three to be slain' (with both auxiliary do and the causati, ve Mv do). Pastor*, Vol. 4.212.10 ,,1460, No. 403) : [They] with here hevedy and fumows langage have ~nd d a y l y do u t t y r lewd and shrewd dalyauns. ' T h e y with their heady and irrascible language have uttered and d~.dy do utter lewd and sharp dallying speeches' (do parallel to have, in code), Paston, Vol. 4.149.37 (1465, No. 585): More playnly than I m a y do w r y t e at thys tyme. 'More plainly than I m a y write at this time' (do after moda'). Chaucer, Cl~'tCs T. 1096: . . . b u t God, of his mercy [ And youre b e n y n g n e fader te:ldrely / H a t h doon you kept. ' b u t God, of his mercy, and y o u r benign father in his tenderness has preserved you' (have - Pe, do - Pp). ~ a u ~ , , ~ v , . Foules I~o: For w~rn that oon encresede ay m y fere, / And with th~.t other gan m y n herte bolde; / T h a t oon me hette, t h a t o t h e r dide me colde. 'For with the one m y fear ever increased, a~d with the other m y heart took cour tge; the one warmed me, the o t a e r chilled me (made me cold ?)' ~gi~ / do / lack of auxiliary apparently all equivalent}. Bevis o/Hamtoun 797 (~s A) : ] )us lae bata:ile gan leste long / Til the time of euesong (MS SN: laus lae t ataile lasted long). 'Thus the battle lasted long until the time of eve lsong.' Cumor Mnn~i 2009 (MS (;6t.) : n, neu liuelad gan he bigin (us Trin. C. dud). 'Hv began a new life' (au:dliary gin beside Mv bigin,.s) S ~ems to be qua d-aspectual, inCieeting immediacy, intertsity, sometimes even ingression. Frequently, especially with ~'A..;'r, howeve,~, it has li':tle farction beyond that of a filler; this is particularly true in poetry, 'where it often seems to be used for rhythmic reascns or to facil;.t~.te rhyme. Do - INF also, especially with PAST,. often has little if any meaning; in some contexts, however, it is cleally asseverative or inten'.~ffying and can possibly be regarded as the overt realization of a u~iversal (?) deep :~tructure elemer..t 'assertion' which at other perio,ts is only covert. By 1he ::eventeenth century gin (now no longer requiring an INF marker since all instances of INF were lost at the end of the ME ,gin - INF 5) Quoted oy the Middle Englis1~ Die:ionary under ginnen 3b.; a N o t h e r n text. T O W A R D A GRAMMAR OF S Y N T A C T I C C H A N G E 19 period) drops out almost cmapletely, leaving no recoverable trace. It l~.rsisted longer with PAST -.han with PRESENt; with PAST it is very common m Spencer, but r a n in Shakespeare and Milton; the latter does not use it at all with PR ~SENT. While use of gin in the late sixteenth ceatury is largely poetic, a few instances of prose use occur, e.g. N~she. Pierce Penilesse 226:16: who awaking his hundred eies at these unexpected tidings ga,~ persue them wheresoever they went. Do (like gin losing I~F at the end of the ME period), on the other hand, has a very interesting history. In Southern English of the sixteenth c e n t a r y , it became m u t u a l l y exclusive with all other au~iliar:z verbs (hence +incoming more generalized, i.e. undergoing simp.ification), s) Relics of the ME usage, however, still persist in those di:dects which allow such constructions as I have done cook and much more common!y ! have ,!one g~.e ,,-.,,., .,-, t. "-".".,,s '-..'.'+~'+:+"'+"'-'~"'+'.',+..t;~,,,.uty do Lecame more and more restricted to negatives, inte+rogatives, erqphatics, and ta~.s where no other auxiliary verb occurred; for S t a n d a r d English, it i,: now totMly predictable. Tills change has involved a shift in function from asseverative to mere tense-carrier, an~:l Lence from a (sometimes) meaningful surface element to an ' e m p t y ' one. It has also brought about simplific~tion of the PS, and of the s f s necessary for interrogative, negative, etc. ; but it has also brought about the addition in the T-rules of a whole new 'do-support' :u,e. The changes in the s f s of interrogative, etc., are simplifications, since both do and V are removed from t h , structural index specifications; e.g. the interrogative shift rule ~_~r (late) ME and ENE ret u i r e s at ieast" SI: X NP Y TEN:~E I~: 2; Lave be do gin V 1 2 3 4 sc: ~ 1 , 3 . 2 , 4 ~) The Ox/ovd English Divtiona,y cites under do III. 31 several six'teenthc e r t u r y Scottish instances of auxiliary do with other auxiliaries, including: 20 E L I Z A B E T H CLOSS T R A U G O T T but for NE we need state only: SI: X NP Y TENSE ([Mhave [ / Z 1 2 \[be _ 3 4 so: --~ 1,3,2,4 As was stated in the pre ¢ious section, according to the view of simplification and elaboratim taken here, sI changes that xesult in generalization are in,,tance,; of simplification, not neces~trily in terms cf the number of elements in a given sI, but in terms of the relationships in the grammar as a whole. By contra,,~;t, increasing restriction in the sI can be considered as an example ,:~felaboration or addition. The result of these restrictions may be fewer morphophonemic and hence fewer surface structure alternatives; but as our emphasis in the whole of this discussion is on change.,~ in ff:ammar, not changes in the surface structure itself, the numbc r of available surface patterns does not concern u~, The inverse relationship between simplification and elaboration of grammatical structures and increases ox decrease in the number of surface patterns available is of con._iderable interest for performance grammars, for the theory of language acquisition, and for assumptions about why langua[e cbanges. Details of this inverse relationship must, however, awa it further studies of changes in competence grammars. A particularly clear case of the increase4 constraints in sI's available for certain sc's can be se..'n in the differences between ENE and N~ restrictive relative clauses. In NE, a subject relative m a y be deleted only if it is followed by be; and if this subject relative is deleted then be faust be too. So we find: I hav,~ a book (which is) yell ~w with age. *I h a w a book is yellow with age. *I have a friend m a y be a gen'.us. Dougl,~s, Aeneis XU.x. 103 (1513): Onto his ceptre thou sail (MODAL) do succeed; Dunbar~ Lam¢,.at lot Maker',',s 49 (159~): He he.¢ (PERFECT)do~e petuously devour The noble Chaucer; Scot. Poems 16th C. II.189 (1578~. And m a n y other fals abusic,n The Paip hes (PERFr~C'f) done invent. T O W A R D A GRAMMAR OF S~¢ NTACTIC C H A N G E 21 In ENE, however, any (surface) subject reDtive couht be deleted, even if ir, troduced by existential and cleft sentence canstructions. Be following such a deleted relative could optionally be deleted, as in the first example below, but more often it was not, as in the second example: Shakespeare. A s you like it II ,i v. 74: Here's a youn g maid with t r a v e l , a uch oppress'd (subject relat~v~ and be deleted) ] And faints for st :cour (s~bject relative deleted). Shakespeare, Mea~sure/or mea, ure II.ii.34: I :have a brother is condemn'd to die (sttbje~:t relative deleted, but not ~.,). Shakest~are, Ring John IV.ii.69: This is the m a a should do the bloody deed (subject relative deleted). Shakespeare, As you like it II[.ii.377: There is a man haur~ts the forest (existential subject deleted). S h a k ~ p e a r e . M~ery u~v.,~s IV.v.iS: It is thine host, thine E p h ~ i a n calls (cleft-complement ~ubject relative deleted). The relative-reduction rule for ENE (and ME) is therefore much more general titan it is for NE. It is even possible that it can be accounted for in th~ ~ame rule as that which allows deletion of object relatives (as in NI:) as well as adverbial relatives (unlike NE), cf: Shakesp.~.axe, A:; you like it III.ii.75: I earn t h a t I eat, get t h a t I wear (obj,~t rele five deleted). Shakespeare, I He#ry VI II.v.5,3: Declare the c a u s e / M y fal:her, Earl of Ca~nbridge ilost his head (adverb relative ~ hy deleted). In that c a~;e, t h s particular instance of restriction in N E might be characterized by addition of a subject deletion rule as opposed to a nc n-subj,.'et deletion rule; whether such an additional rule is necessa.ry or not, additional constraints on the s f s have to t e imposed, specifyinA undeilying relationships, thus leading to elaboration of tt, e st. Additions in sc's may involve the introduction of additional fe, Lture markings, of additional oblige:tory elements, or of extra, al, ernate sc's. Whether the latter can always be treated as part of th~ same vale ol shov, ld be introduced as new rules is not always : k a r ; such deci~:,ons often depend on the complexity of the specific grammar being written. The history of factive nominalizations pro,,,ides a possible example of the addition of alternative so's; in NE we have both: 22 E L I Z A B E T H CLOSS T R A U G O T T T h a t he came was good. I t was good t h a t he came. I~- has been fairly conclusively shown (Lakoff 1965) that such sentences should both be derived *,rom something like It [That he canur] was good with alternative sc's allowing either deletion of It or l~i.extraposition. In oE the c*]y possible fcrm was extraposition (with or without overt hit/daet as complement introducer): Alfred, Or. 194:28: H i t wars 13eh swil~ sweotol p a s t se i h a C r i s t . . . hi~;~ ~ o n e t e n to gescildnesse o m e n d e . ' I t w a s , however, very clear t h a t t~e s a m e C h r i s t . . . s e n t t h e m the ~ , i n as a protection.' *pact se ilca C r i s t . . . hint t~one t e n to gesciidnesse o n s e n d e wars swi~e sweotol. ' T h a t ;~he s a m e C h r i s t . . . s e n t the~,a the r a i n as a ~ cff~¢tion w a s very. clear.' One kind of change that ~ppears not to be readily accounted fl~r in terms of addition or simplification is the change of statt~s from obligatory to optional (cf. the optional deletion of be after relative deletion in ENE, replaced in N~ by obligatory" deletion of be if ~relative deletion occurs), or vice versa. The reason for this is that the ~umber of rules does not change; nothing changes except the availability c~f choice. Once the problem is put in these terms, however, it can l~e seen that a change from optional to obligatory stat~,s is a ~implifiication only insofai as it allows fewer alternatives in t l~e s~.,.rface structure; in terms ~,f tho internal grammar, it is an elaboration as it imposes constraint ~. The preceding discussion of addition ~ o u l d indicate quite clearly that while grammars do often move in the direcdoi~ ~-~fmaximum generaKzation, there is also a very marked counter-tendency toward increased restriction. It has been suggested for phonololV that phonological change characteristically involves a chalige from sporadic -,- conditiom:d -~, unconditioned occurrence (,:f. es~cially Greenberg 1964; Goman 1967). While 'sporadic' occt~rrence has no exact analogue in syntax, it may be loosely equated with optional occurrence in certain stn~ctures at an innovatory stag,~, or in limited registers, as can be illustrated by reference to the use ~f auxiliary do in 1VIEand ENE. But, as do also illustrates, the sequence conditioned unconditioned is by no means necessary or even lik~-ty at a later stage. All kinds of increased conditioning occur frequently in syntax. All that it is possible to do is to establish certain favorable conditions ~t'O~,'ARD A G R A M M A E OF S Y N T A C T I C ( H A N G E 23 for simptificatio~t or elaboration. The :5,'ection theft will be taken cannot, however, be determined. A GRAMMAR OF CHANGF M ~ t of the discu~,~ion in these pages has predicated comparisof o| ,.grammars and some machinery that: would convert grammar A ;.nto grammar B. This has its limitations, although it is a necessary starting place. One of its chief drawbacks is that it turns attention ~way from universals to langnaage-spe.cific problems, and, a practical but im~wtant matter, inevitably entangles one ill the problems of diff~ent;,ating historicM from synchronic order. If a synchronic grammar is a particular subs~:t of universals of synt-actic, semantic, and phonological features, together with~ language-specific rules, we shouM surely wish not just to add individual rifles to some grammar in order to predict change to another state of the same language. We should also hope eventually to establish a theory of change, which itself would be a uriversal grammar of types of changes, including at least some of t;aose suggested in the main part of this paper. This would constitttte a .,,eparate model from synchronic models. Changes applying to particular synchronic grammars would be particular examples of rules from the universal grammar of change~ It would then be possible on the one hand to map changes omo syl~ch~,:mic grammars so as to account for the step-by-step innovations ~hat t<Jok place from period to period and so ultimately for the mutitions that resulted in r,~,lical restructuring. On the other l~and, it woald also be possible to map changes onto sets of strings from variou.~ grammars for the purpose of establishing typologies of chang,~.~, regardh;ss of language fatally, or else for establishing the recur: ent patterns of certain cha'~ges wi:hin one family groap, W~ would then have anal%~es, in th(first case to genetic relati~,nship~" ~nd in the second to 'typologica! r,'lationships', both deriv~ ble from the sam~; universal grammar of chang~. Unive~si~), o/C,~li[ovnia, Depaytmen! o] Enggish, Be~,keiey, (all/.. U 5;.A. Sgar /ord Lniversity, Cc,mmittee on Linguistics, Stan/ord, Call]., U.S.A. 24 ELIZABETH CLOSS TRAUGOTT REFERENCES [For the sake of brevity, the following abbreviation is used: PEGS--Program /or the exchange o/ generative ,p~mmars. Washington, Center for at,plied linguistics.] A:3BOTT, E,. A., 1877. A Sha~espearean grammar. London, Macmillan. A~ Anglo.Saxon Dictionary and Supplemeat, 1954. Ed. J. Bosworth and T. N. Toiler. Oaford, Clarendon Press (re~,r.). ANDF.RSO~, J., 1968. 'Ergat:,.veand nomin,-tive in English', J L 4, I-,32. BActLE., 1962. 'The order of elements in a transformational grammar of German', Lg 2,8,263-6-9. BACQUET, P., 1962. La structure de la phrase verbals d l'dpoque Al/r~dienne. Paris, Les Belles Lettres. BLOOMFIELD, L., 1939. 'Menomini morphophonemics', T C L P 8, 105-15. BowzRs, J. S., 1967. 'Generic sentences in English', (unpub. paper). BOarD, J. C., and J. P. THO~NE, 1968. 'The deep grammax of modal verbs', {to appear- in JL; available as P E G S 31, April, 1968). CltOMSKY, 1Y[., 1965. Aspects o/the theory o/syntax. Cambridge, Mass., MIT Press. CHOMSKY, IV., 1967. 'Some gener~.lproperties of phonological rules', Lg 4,3, t02-28. CHOMSKY, N., and M HALLI,.', 1968. The sound pattern o/English. New York, 7Iarper and Row. CL:~SS, Elizabeth, ; 965. 'Diachronic syntax and generative grammar'. Lg 41, ,102-15. COI.LINGE, N. E., forthcoming. Linguistic models: a h,~torian's use. DAM, J. VAN, 1957. The causal ctause and causal pre#osltlons in Early Old English prose. Groningen, Wolters. ELLEGARD, A., 1953. "rhe auxiliary do; the establishment and regulation of il:s use: in English', Gothenburg Studies in English 2. Stockholm, Almqvist & Wiksell. ENGBLOM, V., 1938. 'On the origin and development of the auxiliary do', Lurid Studies in English 6. Lurid, Sweden, Gleerup. FiL.r~orm, ,~. j., 1966. 'A proposal concerning English prepositions', Georgetown University liIonograph Series on Language and Linguistics 19, 19-33. FILLMOt~E, C. J., 1967. 'On the syntax of preverb.,¢, Gio~sa 1, 91-125. t'Ral~z, W., 1939 Die SFrache Shakespeares in Vers ~nd Pr~sa. Halle/Saale, Niemeyer, 4th ed. FU,'rKE, O., 1922. 'Die Ftigung ginnen mit dem Infinitiv im Mittelenglischen', Engli'.;che Studien 56, 1-27. GOMAI~, R. D., 1967. Toward a unified method o/describing langt, age change (unpub. M.A. thesis, Ohio State Univ. ; available as PEGS 13, Decernbe~, 11967). GaEg~rnEr~6, J. H., 1961. 'Some universals of grammar with particular refer- TOWARD A GRA~tMAR OF S Y N T A C T I C CHANGE 25 ence to the order of meaningful elements' in Univermls o/L.nguage. Ed. J. H. Greenberg. Cambr, dge Mass., MIT Press, 1966, 7~'~-113 2rid ed. GREENBERG, J. H., 1964. 'Language universai,~' i~ Cuwe,~t trends in linguistics 3: ghe~etic~l/oundations. Ed, Th. Sebeok. 's-Graver ~age, Mouton, 1966, 61-112. GREgNBi~RG, J. H., 1966. 'Sy~lchronic and diachronic uni'.ersals in phonology', Lg 42, 508-17. HALtaL M., 1961. 'On the :ole of simplicity ia linguist*: descriptions', Proceedings e/ the S)~mposi~ in applied mathematics t2: ~Ouc,~ure o/ language and its mathematical aspects. Ed. R. Jakobs,.~n. Provi, enee, Rl~ocle Island, American Mathematical Society. 89-94. HALLE, M., 1962. 'Phol~ology in generative grammaJ., Wo~d 18, ,';4-72 (t962), and (slightly ~evised) in The structure o/ lang'~age: readings tn the p~,ilosaphy o/ la,sguagr. Eds. J. A. Fodor and . i . J . Katz. New Jersey, Prentice-Hall, 1964, 334~52. HALLIDAY, 5f, A, K., 1967, 'Notes on transitivity and theme in English', J L 3, 37-81. HO~ZN1GSWALt), H M., 1960. Language change and linguistic reconstruction. Chicago, University of Chicago Press. HOt:mGSWALD, H. M., 1961. 'Are there universals of linguistic change?' in Universals of language. Ed. J. H. Greenberg. Cambridge, Mass, MIT Press, 1966, 30--52. 2nil ed. JI~SPERSEN, O., 1909-49. A Modern English grammar on historical principles, 7 vols. London, Allen and Lrnwin. KAT£, J. J.. and P. M. POSTAL, 1964. An int,,~grated theo:~y o! linguistic descriptdons. Cambridge, Mass., M1T Press. KtPARSKV, P., 1967. ' A p r o p o s de l'histoire de l'accentuation grecque', La#gages 8, 73-93. l~.tr^asgv, P., 1968. 'LingaisUc universals and ! n~mstic change', in Universals in linguistic theory. Eds. E, Bach and R Harm.~ New York: Holt, Rinehart and Winston, 170--202. KLI~A, E. S., 1964a. 'Negation in English', in The struct.,4re o/ language: readings in the. phdosophy ~¢ language. Eds. J. A. FoSor and J. J. Katz. New Jersey: Preaatice-Hall, 1964, 246-323. KLtMA, E. S., ! 964b, Relatedness between grammatical s estems', Lg 40, 1-20. KuRvtowlcz, J., iq62. O n the methoda of internal reconstruction', in Proceedings o/ th~ nin,'h international congress o/ ling ,~ists. Ed, H. Lunt. 's-Gravenhage, Mcuton~ 1964, 9-36. LaKOFF, G., :965. 'On the nature of syntactm irregul trity', Mathematwal linguistics and auto natic translation, Report N o. NSF- 16. Cambridge, Mass., Harvard Computation Laboratory. LAKOFF, G., 1966. 'A note, on negation', Mathematical and automatic translation, Report No. NSF-*.7 Sect]ot~ 111. Camb'ddge, Ma~;s., Harvard Computation Laborato~,. L ~ s o ~ v o z ~ , D. T., 1966 'The syntax of the English expletive " I t " ' , Georgetown University ,~fonograph Series on Language and Linguistics 19, 207-16. d 26 E L I Z A B E T I I CLO.~S TRAUGOTT LEHMA~N, W. P., 1962. Hi~torical linguistics: an int:'oduction. New York, Holt, Rinehart and Winston. ]~[cCAWLeV, J. D., 1967. 'Why auxili~ries axe verbs', (unpub. paper). MCCAWLEY, J. D., 1968. 'Concerning the base component of a transformational grammar', FL A., 243-69. Mi£.tle English Dictionary (1952-). Ed. H. Kurath and S. M. Kuhn. Ann Arbor, Univ. of Michigan Press. ~[oss]~, F., 1938. Histoire de la forme p~riphrastique Street participe pr/.sent en ge¢mam~lue. Paris, Klincksi,~k. Moss~, F., 1949--50. Manual dt l'anglais du moyen-dge: des orig~nes au X I V e s/~de. 2 vols. Paris, Aubier. MUSrANOJA, T. F., 1960. Middle English syntax. Helsinki, Societ6 Philologique. Ox~,~'d English Dictionary, 19,33. Ed. J. A. Murray. Oxford, Clarendon Press. POf.T~L, P. M., 1966. 'On ~,o-called "pronouns" in English', Georgetown l~o~ograph Series on Language and Linguistics 19, 177-206. QuxRs, R. and C. I,. WRENN, 1957. An Old English ~mmaro _London, ~.Vlethuen. 2nd ed. I,~OSJ~tmAUM, P. S. and Dorita LOCHAK, 1966. 'The IBM CORE grammar of 'English', in Specification and utilization of a transformational grammar, Scicatific Report I. Ed. D. Lieberman. Yorktown Heights, N.Y., IBM Corp., Thomas J. ~Natson Research Center. ]~'~oss. J. R., 1967. 'Auxiliaries as main verbs', (unpub. paper). ,C;ALMON, V!ivian, 1%5. 'Sentence structure in colloquial Shakespeaxean English', TransactJ:ons o/~he Phil,dogical Society 105-40. ,';aI'OR'L~A,S., 1965. 'Ordered rules, dialect d~fference~, ~tnd historical processes', L; 41,218-24. ~M1rn, Carlota S., 1%4, 'Determiners aud relative clauses in a generative grammax of English', Lg 40, 37-52. ',~'rocKWELL, R. P., 1964. 'Realism in historical ErLglish phonology', (unpub. paper delivered at LSA Winter Meeting, December, t964). T s ~ R , K. V., 1962. 'Algonquian languages and genetic relationship' in Proceedings of the ninth international congress o/ linguists. Ed. It. Lunt. 's-Gravenhage, Mouton, 1964, 1026-34. TRAU(iOTZr, Elizabeth Closs, 1967. 'Deep and surface structure in Alfredian prose', (ur~pub p~per: available as P E G S 1.4,December, 1967). TR~V¢~oxT, Elizabeth Closs, 1968. 'Old English G r a m m a r I' it, Comput,r exl~riments in transformational grammar, Report AF-33, CS-10& Ed. Jt)y,-e Friedman. Stanford, Stanford Univ. Computer Sciezice Department. TRAUC,OTT, Elizabeth Closs, and J. WATERHOUSE, 1968. 'Already and yet: a sr.pptetive set of aspect-rv arkers ?' (forthcoming in JL ; available as PEGS 3,!~, ~ a y , 1968). UCLA, 19~:~7.A iv Force E~glish Syntax Con/erence, September ~967, Working Fape,,s, (unpub. Los Angeles Univ. of California at Los Angeles, Department of Linguistics). T O W A R D A GRAMMAR OF S Y N T A C T I C C H A ~ G E 27 VISS~R, F. Th., 1963-66. A n historical syntax o/t;w English lang.,~age. 2 vols. ]Leiden, Brill. W~,'.INnc~Icx, U., 1964. 'Explorations in semantic theory' in Current trends in ling~'~istic$ 3: t~oreticag ]oundations. Ed. Th. Sebeok. 's-Gravenhage, Mouton, 1966, 395-477.