1969. “a study in diachronic morphophonemics: the ukrainian prefixes”

24
I lr I I' Reprinted Jrom LANGUAGE Vol. 45, No.4, December, 1969 Printeil in U.S.A. A STUDY IN DIACHROI'{IC N,IORPHOPHOI{IIi\IICS: THE UKRAINIAN PITEF'IXES Hnxxruc Axrnnsnu Haruard Uniuersi,ty Over the last eight centuries the morphophonemic rules relevant to prefixes irr Ukrainian have undergone a number of changes which are here defined and dated relative to one another. The changes are then categorized typologically, and the mechanism of change involved in two of the types is discussed in detail. It is shown that the direction of change in morphophonemics is determined by different synchronic relations between rules and marked lexical items. The changes an&- Iysed permit important inferences about the abstractness of phonological repre- sentations. 1. The purpose of this paper is twofold. Its first aim is to reconstruct the his- torical development of that part of the Ukrairriarr morphophonemic system rvhich is relevant for prefixes and prepositions. Second, the concrete examples furnished by this study will enable us to make some observations on how morpho- phonemic rules change. These observations, as \\'e shall see, have important implications for understanding the nature of morphophonemic rules. 1.1. The changesrvith u'hich rve rvill be concerned l'ould be called analogical by some.We will not use this term, for terms like 'grammatical analogy' as tradi- tionally used are highly ambiguous, and rve need to be specific.In particular, rve nust distinguish between trvo fundamentally different kinds of linguistic change, both of which have traditionally been subsumed under 'grammatical analogy', MORpHopHoNaMrc cHANGE and MoRpHoLoGrcAL cHAr.IGE.l N{orphophonemic ehange is change in a morphophonemic rule or in its application. Since morpho- phonemic rules determine thc realization of morphemes, a morphophonemic charrge u'ill lead to change in the realization of morphemes without any change in their basic forms. \Iorphological change, bv contrast, is change in those rules of a language rvhich assign phonological form (segments, distinctive feature specifications) to elements of grammatical meaning. Since morphological change concerns meaningful elements, anv morphological change involves relations be- trveen linguistic signs. Morphophonernic change, on the other hand, concerns the relation betrveen contextual variants of the s&me linguistic sign(s). This distinc- tion, it should be noted, is not reflected in the traditionally recogn\zed,but sterile, subelassification of 'analogical change' into 'material' and 'formal'. It is sympto- matic of the degree of clarity hitherto attained in the study of 'grammatical analogl" that the obviously basic distinction bett'een morphological change arrd morphophonemic change is nolvhere made plain in the li'ritings of such modern studerrtsof analogy as Bulaxovs'kyj (1940, 1956a),Kurylorvicz (1960), \'Iafrczak (1958),Schmidt (1955),and Zirmunskij (1954).' I 'Morphophonemic ehange' in the usage of some lingrrists refers to the effects of morpho- phonemic rules, i.e. to sylchronic morphophonemic alternations. I will use the word 'change'in this and other phrases, like 'phonological ehange' and 'morphological change', exclusively in the meaning 'historical change'. ? As a eharaeteristic example of the traditional corrfrrsion of morphophonemics with I t,r

Upload: ucla

Post on 14-May-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

Il rI

I'

Reprinted Jrom LANGUAGEVol. 45, No.4, December, 1969

Printeil in U.S.A.

A STUDY IN DIACHROI'{IC N,IORPHOPHOI{IIi\IICS: THEUKRAINIAN PITEF'IXES

Hnxxruc Axrnnsnu

Haruard Uniuersi,ty

Over the last eight centuries the morphophonemic rules relevant to prefixes irrUkrainian have undergone a number of changes which are here defined and datedrelative to one another. The changes are then categorized typologically, and themechanism of change involved in two of the types is discussed in detail. It is

shown that the direction of change in morphophonemics is determined by different

synchronic relations between rules and marked lexical items. The changes an&-Iysed permit important inferences about the abstractness of phonological repre-

sentat ions.

1. The purpose of this paper is twofold. Its first aim is to reconstruct the his-torical development of that part of the Ukrairriarr morphophonemic systemrvhich is relevant for prefixes and prepositions. Second, the concrete examplesfurnished by this study will enable us to make some observations on how morpho-phonemic rules change. These observations, as \\'e shall see, have importantimplications for understanding the nature of morphophonemic rules.

1.1. The changes rvith u'hich rve rvill be concerned l'ould be called analogicalby some. We will not use this term, for terms like 'grammatical analogy' as tradi-tionally used are highly ambiguous, and rve need to be specific. In particular, rvenust distinguish between trvo fundamentally different kinds of linguistic change,both of which have traditionally been subsumed under 'grammatical analogy',MORpHopHoNaMrc cHANGE and MoRpHoLoGrcAL cHAr.IGE.l N{orphophonemicehange is change in a morphophonemic rule or in its application. Since morpho-phonemic rules determine thc realization of morphemes, a morphophonemiccharrge u'ill lead to change in the realization of morphemes without any changein their basic forms. \Iorphological change, bv contrast, is change in those rulesof a language rvhich assign phonological form (segments, distinctive featurespecifications) to elements of grammatical meaning. Since morphological changeconcerns meaningful elements, anv morphological change involves relations be-trveen linguistic signs. Morphophonernic change, on the other hand, concerns therelation betrveen contextual variants of the s&me linguistic sign(s). This distinc-tion, it should be noted, is not reflected in the traditionally recogn\zed, but sterile,subelassification of 'analogical change' into 'material' and 'formal'. It is sympto-matic of the degree of clarity hitherto attained in the study of 'grammatical

analogl" that the obviously basic distinction bett'een morphological change arrdmorphophonemic change is nolvhere made plain in the li'ritings of such modernstuderrts of analogy as Bulaxovs'kyj (1940, 1956a), Kurylorvicz (1960), \'Iafrczak(1958), Schmidt (1955), and Zirmunskij (1954).'

I 'Morphophonemic ehange' in the usage of some lingrrists refers to the effects of morpho-

phonemic rules, i.e. to sylchronic morphophonemic alternations. I will use the word'change'in this and other phrases, l ike 'phonological ehange' and 'morphological change',

exclusively in the meaning 'historical change'.? As a eharaeter is t ic example of the t radi t ional corr f r rs ion of morphophonemics wi th

I

t,r

808 T,ANGUAGE, VOLUtrlt 45', NUMBER 4 (1969)

There is no doubt that the faiiurc of traditional historical linguistics to dis-tinguish betu-een morphological and morphophonemic change is responsillie forthe fact that there has been no study of morphophonemics aimed at establishingits general principles of change. It is hoped that the results of this investigationrvill be a contribution torvard that end. It seems that the study of morphopho-nemics as an autonomous linguistic subsystem is particularly necessary at thepresent time u'hen many investigators tend to consider morphophonemic rulesas simply part of the phonology of a language. The study of change in morphopho-nemics is especially important, for while in a static synchronic description it maybe possible to ignore essential differenees between morphophonemic and phono-logical rules, no iinguistic description which aims to account for the dynamicaspects of language-whether in synchrony or diachrony-can disregard thedifferent part-whole relations of morphophonemic rules and phonological ruleswithin a linguistic system. These different relations are manifested in essentiallydifferent kinds of change. We will not be able to discuss the differences bctrveenmorphophonemic and phonological change in any detail in this paper; but, a-s rveshall see, the results of our investigation clearly demonstrate the neecl to dis-criminate between phonology and morphophonemics.

1.2. The subject chosen for our study-the history of the Ukrainian prefixes-has not been treated systematically before. Historical grammars of lJkrainianmerel5' acknowledge that the prefixes have undergone analogical change.'|

'I'rvo

articles by Zovtobrjux (1958, 1962) are in part specifically devoted to sorne ofthese changes, but Zovtobrjux's approach is thoroughly traditional. He does nottry to establish how the morphophonemic system has changed, but insteiirl cort-jectures how the shape of individual lexemes may have been influenced br" otherlexemes. The unmitigated atomism of this approach gives no hint tliat thechanges in question are general in nature (i.e. involve rules), and hence effectivelyobscures-for the investigator as rvell as for his reader-the fact that thesechanges are tvpologically very different.a Some questions of detail coltcerningthe history of ukrainian prefixes have been discussed elservhere (Andersen 1969a,1969b). The conclusions reached there rvill be utilized belorv.

morphology, one can ment ion Kurylowicz 's not ion of 'morphbme compos6' , meaning a

morpheme plus the effects of a morphophonemic rule applying in an environment defined

with reference to that morpheme, e.g.-e p lus umlaut in the plural of German Bau,rn ' t ree ' ,

Briume (1960:70). In his most recent work, Kurylowicz appears to argue against this impor-

tant notion in his theory of analogy (though without rnentioning it), recognizing eorrectly

that umlaut is 'semantically void' (1968:75,79). Flere, however, he proposes to define mor-

phophonemics as the study of 'semantically void' morphs (p. 80) with the exception of'structures dominated by ... phonological factors' (79). His definition ignores the fact that

morphophonemics does not specifically deal with invariant units of form, and he does not

try to justify the exclusion of phonologically conditioned alternations from morphopho-

nemics. His conclusion that morphophonemics belongs to morphology is clcarly consistent

with his initial assumptit-rns, l:ut sheds no light either on the nattrre of morphophonemics

or on the relation between morphophorremics and morphology.3 See , fo r ins tance ,Hruns ' ky j &Kova l ' ov t r941 :93 , Bezpa l ' koe ta l . 1957 :150-1 , Bevzenko

1960: 391, Nledvedjev 1964: 115.a Zovtobrjux 1958 deals with Charrges 2a and 3 of the typological classification below

($5.2).

DIACHRONIC X'IOIIPHOI'}HONEN{ICS: 'f tIIJ UKRAINIAN PIiEFIXES 809

1.3. The subject of our study is at orrce u,ider and narrorvel than the referenceto 'the t1krairiian prefixes' of the titie suggests. In the first place, our discussionrvill concern not only prefixes, but also proclitic prepositions. These two mor-pheme classes are treated identically, morphophoiremically and phonologically,in Oid Ilussian. We u'ill refer to both as 'prefixes', mainly because a number of themortrlhophonemic rules rn'e lvill discuss, quite independently of other changes,harrc ceased to be applicabie to prepositions, but continue to appiy to prefixes inmodern Ukrainian (see $4.10 below). In the few cases rvhere reference to preposi-tions rnust be excluded, we will speak of 'prefixes sensu stricto'.

In the second place, not all the prefixes of Ukrainian u,ill be discussed. Prefixesu'hich clo not end in a consonant, in b, e, or o in Old Russian are of no interestfrom the point of vierv of diachronic morphophonemics; rve have rrothing to sayabout srueh prefixes as n,a'ort ', pri 'at 'r n,'11,L', ea 'behind'. The Old Russian prepo-sitiorrs ita 'to' and aw'along' have fallen into disuse. They and the prefix u'i, 'ir.',

rvhich has merged rvith u-both are represented by u in modern lJkrainian-rvillaiso renrain outside our discussion.s

l,'inailv it should be mentioned that the Old Russian prefixes sa 'off; rvith' and

.7oe 'out of'have merged into a single prefix, modern Ukr. e (cf . Andersen 1969b).

This morphological change, as u'ell as the merger of uz, and u mentioned above,is independent of the morphophonemic clianges rve will discuss belolv.

1.4. In our discussion of the historical development, rve wili refer to threestages in the development of Ukrainian: Old Russian (ORu.), Early Ukrainian(trL,lir.), and Contemporary Standard Ukrainian (CSU). The last is the officiallanguage of the URSR as described in normative grammars (e.g. Bulaxovs'liyj1951) Lind dictionaries (e.g. Kyrydenko 1953-63).6 Only in a feu'instances will rvehavc occasion to cite forms from other varieties of contemporary Ukrainian.B.r, 'Old Russian' \\'e mean the pre-Ilkrainian dialects of Old Russian at a stagestrortly before the fall of the jers (cf. $3 below), i.e. around 1100.? We use theterm 'INarly Ukrainian' to designate the same dialects immediately after the fallof thc jers (second half of the twelfth century).8

2. We begin by surveving the rules, phonological as u'ell as morphophonemic,

6 There are still prefixal formations and fixed prepositiorial phrases in which ORu. aa isrepresented by uui, rather than z. These residual forms are of some interest in connectionwith the discussion in $4.8; cf. fn. 21.

6 In citing CSU forms we use a broad transcription which is slightly less explicit than theCSU orthography. In this way we avoid the disadvantages of a transliteration and can,among other thiugs, represent similar consonant clusters in a parallel fashion (e.g. bj-ut''beat , ' , l . j -ut ' 'povr ' , phonet ical ly [b jut ' ] , [ ' l 'u t ' ] , in t ransl i terat ion bj 'ut ' , I l ju t ' ) and insert

hyphens at morpheme boundaries, which the orthography does not permit. Where we needto distinguish explicitl5r between underlying representation and phonetic realization, theformer will be enclosed in braces, the latter in square brackets.

7 Old Russian forms will be cited in a transcription about as narrow as the ORu. graphic

system, but differing from the latter in being consistent. All but a few of the ORu.

lexemes cited are attested in early East Slavic texts (cf . Sreznevskij 1958, 1968); the few

that are not are marked with an asterisk, except for prefixed lexemes whose stems are at-

tested with some prefix other than the oue cited.8 Early Ukrainian forms are all hypothetical, and will therefore not be starred. They will

be given in a tra,nscription commensurate with that used for Old Russian forms.

810 LANGUAGE, VOLUI'IE 45, NUMBER I (1969)

which affected the reaiization of prefixes in Old Russian before the fall of the jers.

There are five rules to be considered.2.1. OesTRUENT DELEtroN. At morpheme boundaries, stops are droptrted be-

fore any obstruent or nasal; {t}, {d} are dropped before {l} ; and {.}, tz} before

{s}, {z} , {S}.A. a consequence of th is rule, {ob} 'about ' is posi t ional ly real ized

as [o]; and \juul'out of', {bez} 'without',

\roz} 'aper,rt 'as

[jr ' ], [be], [ro] respec-

tivelv.Originally this rvas a phonological mle applied to all obstruents in svllable-final

position; but at the stage of Old Russian rvith which we are dealing, it had evi-

dently lost its purely phonologic:rl character and had become a morphophonemic

rule. This is clear from the fact that types of obstruent clustels rvhich had previ-

ously not been admitted were now possible: e.g., fricative * fricative at prefix

boundaries, e.g. .jas-rodi,-ti for \jsz-xodi-ti | 'go out'; and stop * fricative iri loan

rvords, e.g. Oleksandrt.2.2. ConruoN Sr,avrc pARAGocn. Before initial obstruent, ttasal, or {l } , a para-

gogic {r} is added to {ot} (cf. Andersen 1969a). This rule applied before Ob-

struent Deletion, and, as a consequence, {ob} and {ot} are distinct in all environ-

ments; cf. (I) belorv.Historically, the introduction of the Common Slavic Paragoge rule followed

the loss of syllable-final obstruents, rvhich had affected {ob} and {ot} equally.

The last phase of the establishment of the Paragoge rule is actually documented

by Old Church Slavonic, rvhere, in a ferv prefixal formations, the archaic o- f.or

{ot} is attested beside the younger, regular otr'-.s

Note: in a few lexemes, CSU has preserved initial o- for *ol, showing that the items inquestion were lexicalized before the Common Slavic Paragoge, e.g. ostauytE 'leave behind',oskiloh'chip'.

2.3. OssrRuENT-pLUS-oBSTRUENT sANDHI. Obstruents become tense before a

tense obstruent, lax before a lax obstruent.l0 Because of this rule, {j"r}, {bez},

{roz} are positionally realized as [jrs], [bes], [ros].I1 origin, this is a constraint, inherited from Indo-European (cf. iWeillet

1964:132) .2.4. OgsTRUENT-pLUS-NoN-oBSTRUENT sANDHI. Before initial vowel or sono-

rant, prefix-finai obstruents are lax. Only the prefix {ot} was afrected by this

rule, being realized as [od] before initial non-obstruents (cf. Andersen 1969a,

1969b). Contrast the realizations of {ob} and {ot} below.

I. [o], [otr] before obstruent or nasal (cf. $$1.1-2);lobl, [ott] before {1} (cf' $2'2);[ob], [od] before vowel, gl ide, or {r}.

Note: in a few lexicalized composita, where the prefix boundary had been obliterated,*ol remained unaffected by this constraint; cf . CSU otruby 'bran', otuorytE 'open'.

2.b. CoypACTNESS ASSrMrLArroN. Sibilants become palatal before palatal

e Diels (1g32:123) discusses the Old Church Slavonic facts. For the attestation in Old

Russian, see Sreznevskij 1958, s'w.r0 Apparently, in the Old Russian dialects from which Ukrainian has developed, tense-

,r"r., oo1 voicing, was phonemic in the obstruent system (cf' Andersen 1969b)'

DIACI{RONIC X{ORPHOPHONITNIICS: THE UKRAINIAN PRE}'IXES 811

consonants. As a consequence of this rule, {jur}, {bez}, {roz} are positionallr '

realized as [jrZ], [beZ], [roZ] as rvell as [juS], [beS], [ro5].2.6. In short, the Old Russian prefixes were subject to the follou'ing rules, ap-

plied in this order: a. Common Slavic Paragoge, b. Obstruent Deletion, c. Ob-

struent-plus-Obstruent Sandhi, d. Obstruent-plus-l{on-obstruent Sandhi, e.

Compactness Assimilation. Of these, the first trvo were morphophonemic rules,

applicable only in specific morphologically defined enl'ironments; the rest \n'ere

phonological rules-neutralization rules, to be precise.

In addition to these rules, the Old Russian prefixes were of course subject to

various phonological rules which specified subphonemic features. Among these

rvas the rule that specified the realization of the jers, {r} and {t}, which in some

environments (termed 'weak position') could be elided, but in other environ-

ments (termed 'strong position') were pronounced as lax diffuse non-compact

vorvels. Also included here was the rule that specified the realization of the non-

diffuse non-compact vowets {e} and {o}, rvhich in rveak position were realized

as lax [e] and [o], but in strong position as tense [6] and [6].Strong and weak position were defined slightly differently for the jers and for

{e } and {o } . Leaving certain details aside, \ve can say that jers rvere weak in all

environments except that, rvhen they occurred in neighboring syllables, eYery

second jer, counting from the last, r,vas strong. We rvill define one more strong

position belorv. {e} and {o} were strong only when the immediately follorving

s-vllable contained a weak jer, but even here they were weak under certain condi-

tions (see Bulaxovs'kyj 1956b:49-58 et passim).

3. In the first half of the twelfth century, the phonological system was re-

structured. The u'eak jers rvere lost. The strong jers tvere identified rvith weak {e}and lo)- the1'cont inuecl as EII l i r . { " } and {o}. The strong {e} and {o} con-

tinued as EUkr. 16) and {6}. In Earlv Ukrairrian, as still in the archaic North

Ukrainian dialects, t6) and {6} rvere opposed to {e} and {o} as tense to lax.

The CSLT ref lexes of t r I lkr . {e} ancl {o} are {e} and {o};EIJkr. {6} and {6} are

reflected as CSII {i}-but in rvord-initial position, {6} has changed to CSII {vi}.All these changes are relevant for the prefixes, for not only did the realizations

of the latter change, but these realizations had to be produced in part by new

rules operating on new basic forms.Before we proceed to examine the nerv rules to which the fall of the jers gave

ri-qe, let us briefly mention the fate of some of the Old Russian rules stated above.

3.1. O3sTRUENT DELETToN. As a consequence of the fall of the jers, the number

of prefixes ending in an obstruent more than doubled; but the Obstruent Deletion

rule rvas appiicable only to those rvhich had originally ended in an obstruent.

Since the fall of the jers, the rule has ceased to be generally applicable even to

them. Thus, {ob} may now be realized as [ob] before any obstruent or nasal, andt,"l , lbez\ , troz ] retain their final [z] before sibilant, as below.ll

rr Verbs are cited in the form which contains the basic stem: the infinitive (in -tE), thethe 3pl. present tense (in -ut'), or, rarely, the singular imperative (in -fl). Here and in the

fol lowing, 'etc. 'at the end of a set of examples indicates that addi t ional examples can

easily be suppliecl from any alphabetical dictionary of Ukrainian. Where this is not the

case, I will try to inclr-rde all pertinent examples.

812 LANGUAGE, \TOLUME 45, NUMBER 4 (1969)

II . {ob}: ob-plet-ut ' ' fence in' , ob-trusy-ty'shake off ' , ob-kury-ty , f .umigate', ob-smoly-ly'tat', ob-zoloty-ty'gild', ob-\yj-ut' 'hem', ob-Eary-ty 'scorch', ob-ci,Iuua-ty ,cover with kissesi,ob-t.11sly-ty 'calculate'

, ob-nyzy-ty ,lower', ob-mouy-ty ,slander', etc.lzl : z sad,u'from the garden', z-zatyl, ja,from behind,, etc.{bez} : bez-son ' ja ' insomnia ' , bez-zuby j ,booth less ' , e tc .lrozl: roz-sudolt'judgement', roz-zuj-ut' ,take shoes off', etc.Note: there &re many lexemes which preserve [o] for {ob} before initial obstruent or

nasal. The synchronic situation is instructive &s an example of how tenacious a morpho-phonemic rule can be iong after it has ceased to be phonologically motivated-in this caseabout 800 years-and long after the domain of the rule has come to be definable only withreference to specific individual morphemes. A small sample will show the complexity ofthe synchronic distribution of [o] and [ob] for {ob}. Kyrydenko contains 166 entries withlobl before initial {h}, represerrting a total of 47 stems.12 The dictionary recognizes only o-(the historically correct realization) before nineteen of these, prefers o- to ob- before three(by cross-referencing), eonsiders o- and ob- equivalent before six stems (giving them sepa-rate entries without cross references), prefers ob- to o- before two (by cross-referencing),and reeognizes only ob- before eleven stems. The remaining six stems have o- in some mean-ings, ob- in others.

A few residual forms have [ro] for {roz}, e.g. ro-sil 'broth' (in the current orthographyrozsil, ORu. ro-solz,), ro-s-lto,za-t'y 'tell' (spelled rozkazaty, ORu. ro-st-kazo-ti, cf . Ru. ros-skazat ') .

3.2. OrsrRUENr-rLUS-oBSTRUENT sANDrrr. After the fall of the jers, the ner,r.obstruent clusters became subject to the sequential constraint 'lax before lax,,but the Old Russian constraint 'TnlrsE-BEFoRn-Tnxsn'

was not extended to thenew obstruent clusters. l\'faintained for some time as a morphophonemic rule, theTense-Before-Tense rule was eventually eliminated rvith the effect that under-lying sequences of lax * tense obstruent re-appeared phonetically. For instance,corresponding to the ORu. infinitives aesti, from ued-uta 'iead', and uesti, fromuez-utb'convey', CSU distinguishes uesty (ued-ut') and uezty (uez-ut'). Similarly,the prefixes {z}, {bez}, {roz} are pronounced with [z] before tenuis, e.g.z sad,u,bez-sonnyj, roz-sudok with [zs] (see further Andersen lg6gb).

3.3. OnsrRUENI-rLUS-NoN-oBSTRUENT sANDHr. After the fall of the jers andthe introduction of the 'lax before lax' constraint, the positional. rcalizations ofthe Old Russian prefix {ot} paralleled those of lz}, {bez}, {roz}:

III. Before tense obstruent: [0t], [bes];before lax obstruent: [6d], [bez];before non-obstruent: [od], [bez].

It I'as re-interpreted as {6d } , and the Obstruent-plus-Non-obstruent sandhi nrlelapsed, since it did not apply to any other morphemes. As long as the Tense-be-fore-Tense rule was maintained in the grammar (cf. $3.2), {6d} could be realizedn'ith a final [t] before tense obstruent, just as lz], {bez}, {roz} were realized n'itha final [s] in this environment. But when the Tense-before-Tense rule was elimi-nated, {6d} came to be realized as [dd] before tenuis. In CSU it is consistentlvpronounced [vid] before tenuis.

4. We can now turn to the alternations created by the fall of the jers and thercstructuring of the vorvel system, and to the new morphophonemic rules thatwere established to produce these alternations.

12 I use the term 'stem' to denote the corrstitr-rent that immediately follows a prefix.

DIACFIRONIC MORPHOP}IONEMICS: TIIE UKRAINIAN PREFIXES 813

4.1. IJrnarNrAN rARAGOGE. Before stems whose first syllabic had been a weali

jer, thefinal jer of ORu. {nadt} 'above',

{ot} 'from' (when realized [otr], cf. $2.2),

{podr} 'under', and {sr}

'with'had been instrong position, andhad changed to

EUkr. {o}. Elsewhere it had dropped.In trying to reconstruct how this 'mobile o' was produced at the stage immedi-

ately follorving the fall of the jers, we must consider four possibilities. It could

have been analysed (1) as part of the prefixes in question, thus setting these

pref ixes apart f rom El lkr . \z\ , lbez| , { roz}, {ob};or (2) as part of the stems in

question, distinguishing them from all other stems. hr either case a morpho-

phonemic rule would delete the {o} in the appropriate environments. Alterna-

tively, it could have been interpreted (3) as a paragogic vowel, added by a

morphophonemic rule to certain prefixes ending in a consonant before certain

stems; or (4) as a prothetic vorvel, added to certain stems after certain prefixes.

We have to discount the first trvo of these possibilities, for in the later develop-

ment the 'mobile o' came to occur both after the prefixes that originally did not

end in a jer, and before stems rvhich originally did not have a weak jer as the first

svllabic. Elements that serve to distinguish morphemes-&s this 'mobile o' would

have-s,hether they be segments or distinctive features, eannot be transferred

from one morpheme to another. If rve accept either of the first trvo possibilities,

ive cannot explain the spread of the 'mobile o'.

If, on the other hand, s-e posit a morphophonemic rule which assigned a

paragogic or prothetie {o } after prefixes ending in a consonant (but with certain

exceptions) before certain stems, then the changes in the occtlrrence of the'rnobile o' can be accounted for as changes in the domain of that rule-and

ulderstood as simplifications of the grammar. We rvill see belou' rvhy the lo )assigned by this rule must be considered paragogic rather than prothetic (cf. note

1g). Let us no\v examine horv the occurrence of the paragogic {o} changed. (The

ehange of paragogic {o} to ti} rvil l be discussed in $4.8')4.2, Inits initial form, as was pointed out above, the Paragoge rule (as we shall

call the lJkrainian paragoge rule from norv on) did not apply after lzl, \bezl,

{roz } , lob } . In CSU, holvever, it applies v'ithout exception after all prefixes end-

i1g in a consonant before all stems rvith which paragoge is regular (see below).

IV. lobl: obi-bj-uC"nail arouncl' (ORu. o-btj-att), obi-bra-ty'choose' (ORu. o-bura-ti),obi-uj-ut','wrap up' (ORu. ob-r,j-tita),rs obi-hna-ty 'otttstrip' (ORu. o-gtna-ti), obi-hnu-l'y,bend' (ORu. o-gz,nu-t'i), obi-ilra-t,11'pull off' (OR,u. o-dura-ti,), obi-\n-u,t' 'finish harvesting'(ORu. o-Etn-rita), etc.

lrozl: rozi-bj-ud''break' (ORg. roz-brj-t i,tu), rozi-bra-tg'take apart' (ORu. roz-bara-t' i),rozi.-uj-ut' 'unwrap' (ORu. roz-uuj-t)ta), rozi-hna-ly 'disperse' (ORu. roz-gz'na-ti), rozi-hnu'tg ,rrnbend, (ORu. roz-grnu-ti), rozi,-(lra-ty'rend' (ORu. roz-dwa-ti), etc.

)iote: the ORu. prefix {jsz} oniy very rarely functioned as a prefix sensu stricto. Noworrll- such fixed prepositional phrases as zo us'ir 'of all ' (ORu. juz uusdrn) or the adverb

zorn'i ,from without' (ORu. jt z uxnd), whence zoun'i,En'' i j 'external', show paragoge with the

or ig inal l jsz l . I t is impossib le to te l l whether paragoge in these cases preceded the merger

o f ORu. { j sz l and {ss } i n Ear l y Uk ra in ian (c f . $1 .3 ) .

13 In Old Rnssian, {vf was el ided af ter {b} . Hence, {ob-br j -utr } and {ob-vsj -utr } were

realized iclentically as obtjtitu. The introduction of paragoge after {ob} has disambiguated

them, bl t one of the mea,nings sr ig inal ly carr ied by {ob-vsj -uts} , v iz . 'deck, wal lpaper ' ,

is now carr ied by obi-b j -ut ' .

,s14 LANGUAGII, VOLUME 45, NUMBER 4 (1969)

{bez} functious as a prefix sensu stricto in nominal formations, otherwise only as a prep-osition. A single sttbstantive, bezo-dn'o 'abyss' (ORu. bez-d,ana), and fixed prepositionalphrases hke bezo us'eo 'without all', bezo us'akoho 'without any kind of' (ORu. bez uasdrr,,bez uusriltogo) suggest that the Paragoge rule once was as regularly applicable to {bez} asto other prefixes ending in a consonant.

The ORu. pref ix {ot} did not have a paragogic jer before init ial {r} (cf. $2.4).But asuidi-rua-l 'y ' tear off ' (ORu. ot-rua-t i , i .e. [odrrvati ]) shows, here also the Paragoge rule hascome to apply.14

4.3. The Paragoge rule, as \vas mentioned earlier, has come to apply beforestems which did not originally have a weak jer as the first syllabic.

In the first place, a paragogic vowel is added to the prefix {z} (oRu. {r"})before stems beginning rvith {s } follou'ed by another consonant:

Y. zi-skoty-ly'jump off' (ORu. sz'-stcoli-ti), zi-skouznzr-fy'slide down' (OIlu. sz,-shalznu-ti), zi-star'ij-ut' 'age' (ORu. sz'-stard.j-titu), zi-str'ily-tg 'shoot dowrr' (ORu. st-strdli-ti),zi-struha-ty'plane off' (ORu. st -struga-ti., sr,-strz,ga-ti).

It is interestirrg to rtote that {bez } and { roz } do not undergo paragoge beforethese stems. The phonological conditioning of this extension of paragoge, andthe fact that it is limited to \z), rvhich originallv ended in a jer, indicate that u,ehave here not the result of a morphophonemic change, but the effects of a pho-netic development. We can interpret it this rvay: in Old Russian, a jer was strongwhen preceded by an initial sibilant and follou'ed by sibilant * consonant.

4.4. But quite apart from the special case just discussed, the number of stemsbefore tvhich all prefixes receive a paragogic vowel has been increased.

The Paragoge rule regularly applies to all prefixes ending in a consonant beforethe stems belorv, u'hose first syilabic in Old Russian was a weak jer. The stems inVI presented a Y-0 alternation rvhere formerly other vowels had alternated. u'iththe weak jer; the stems in VII did not.

Yr. bj-ut ' 'beat ' , bra-ty ' take', uj-ut ' ' twist ' , hna-ty 'chase', hnu,-ty 'bend', d,ny-ty 'puf abottom in' , d'ra-ty' te&r' , drnu-ty'breathe', ida-ty'wait ' , Enr-,ut"press', En-u{ 'reap', zua-111

'cal l ' , jd-ut"walk' , jm-u,t ' , take', l j -ut, ,pour,, mknu-ty,close,, mn-lt t t ,rumple', mr-ut,,d ie ' , p j -u t ' 'd r ink ' , pn-ut ' ' fas ten ' , pr -u t ' ,push ' , pra- ty , launder , , prnu- ty ,k ick ' , rua- ty' tear' , sla' ty 'send', st la-Ly 'spread', spa-ty 'sheep', ssa-ty 'suck', stnu-ty 'dty up', tr-ut ',rub', tka-ty 'w'eave', tknu-ty , touch,, tn-utt ,cut,.

YI I .bha- ty ' fo ld ' , dba- ty ,heed, , zdr , i j -u t , ,see, , z ly - ty ,anger ' , I ,Zy- ty ,grow l ight , , ml , i j -'ut ' ' faint ' , msty-ty 'avenge', m\a-ty 'rLlsh', pr ' i j - tr t ' 'perspire' , psuua-ty'spoi l ' , rnu-ty'drown', t l ' i j -ut ' 'smolder' .

The Paragoge rule carr never h.ave referred to the original shared characteristicof these stems-the fact that they had once contained a rveak jer-for once thejers rvere lost, this shared property was a thing of the past. In the stens listed inVI, V,-'fl alternations were found-and still are-betn'een the infinitive stem

la It may be noted that the regular development of ORu. -ra- in Ukrainian (cf. ORu.rtuati, CSU ruaty, but after consonant ORu. krbuaub, CSTJ hryuauyj 'bloody') would havegiverr rise to a ttnique alternation in the stem of ruaty, conditioned by the prefix-final seg-ment, e.g. na-raa-fy 'pluck',but *aib-ryaa-ty'pluck off ' (ORu. ob-rz,ua-t i , ; the correspondingimperfective verb, ORu ob-ryua-ti, rs ob-ryua-ty / ob-rguaj-ut'). The introduction of paragogeafter {ob}, {6d} , lroz} tttay have been responsible for the loss of this alternation (henceobi -rua-ty like na-rua-ty) .

DIACHRONIC NIOITPHOPHONITMICS: TFIE UKRAINIAN PITITFIXES i i15

arlcl tire iiresent stem, and between a perfective and the correstr)orldirrg irnperfec-tive verrb, as beiou,.

YIIL Pef i . z-by- t 'y / z i -b j -u l ' , impf. z-byuaj-u. t "knock o{ f ' (OI lu. sa-bi - lz f sal t t j - t ' i ta ,

sa-bi r ;a j - i tu) ; p i r t i , -bra- tV / p id-ber-ut ' , p id-byraj -ut ' 'p ick up' (ORu. podz,-bva- l i / podr-

ber-ult,, podz,-bi,raj-ritr); nad;i-hnu-tA / nadi-hn-Ltt', nad-ltynaj-zl' 'bend slightly' (ORu.

nada-gz,nu-ti / nadz'-gz,n-uta, nadz,-gybaj-tita); 16 u'id-mer-ty / uidi'mr-ttt ', u'id-myraj-ut' 'die

awa\r' (ORu. oto-mere-ti f otr,-m,ar-uto, oto-ntiraj -rita).

But in each such instance, all stem alternants could-and carl-be derived onlyfrom a basic stem containing the 'zero vo\yel' (c.g. {bj}

'beat', {bra}

'take', {hnu}

'bend', {mr}

'die').tu The stems listed in VIi did not present Y-A alternations.The only thing ali the stems in question had in common was that they beganrvith eonsonant clusters.

This common characteristic has evidently been abstracted as the conditioning

factor for paragoge. As a corrsequence, the Paragoge rule has become applicable

before stems beginning n'ith a consonant cluster which have never contained au-eak jer. This accounts for the fact that a paragogic vorvel is assigned before the

infinitive stern of {pn} 'spread' and {*t }

'rumple' (IX), rvhere the clusters of

labial + -vod have developed from Early Ukrainian sharped labials; and it ac-

counts for paragoge before the stems represented in X.

IX. t, idi-pja-sty / uidi-pn-u,t ' ' trnfasterr' (OItu. ott,-pri-t i / otu-pan-zrlo); similarly zi-,obi-, rozi-pja-sty; zi-nia-ty f zi-mn-ut"cnrmple' (ORu. sz'-nrci-t i f st-rntn-ult); similarlypidi - . obi - , rozi ' rn ja- ty.

\ . in Uj-ut ' ' tc t t ' r hraj - t t t ' 'p lay ' r hr ' i j -ut"warm', znai :v l "know' , l ; l 'ad-ut ' 'p t l t ' , krad,-ut '

' s tea l ' , p r ' ad -u t ' ' sp in ' , sp i j -u , t ' ' be i n t ime ' .

Paragoge has not been established in a,ll prefixed forms of these stems. For ex-

ample, as seen in \I, Iil'r5'denlio (a) does not record paragoge in some forms, (b)

rnarks paragoge as 'rare' in some, (c) prefcrs the absence of paragoge to its pres-

enee in some b1- cross-referencing, (d) prefers paragoge to its absence in some, (e)

marks the absence of paragoge as 'rare' in one verb, a,nd, (f) gives only forms with

paregoge in some stems.

\ i . (a) p id- , roz-hrauaj 'y1t ; roz-znauaj ; t , i t l - , z- , nad-, ob-, p id- , roz-k ladaj-ut ' , a id ' , z- ,

na r l - ,ob - ,p id - , roz -k lad , -u t ' ; a id , - ,ob - ,p id - , roz -k rada j -u t ' , u i c l - , p id -k rad 'u t ' ; u id , ' , ob -p r ' ada i -

t t t ' , t ' id- , roz-pr ' ad-ut ' .(bt ob ( i ) - , roz( i ) -k lad-ut ' .(e ) ob (i ) -, r oz (i) -kl ad, -ut' ; ob (i) -, r oz (i) -krad -u t' .

(d) z, ic l ( i ) - , z( i ) - , ob( i ) -hraaaj- t t t ' .(e) ob (i) -hrauai'111' .

rb I t is i r r terest ing to note that the stem of ORu. -gybaj-utu has been made over as -hynaj-

rr1'. The lerr form shows unequivocally that imperfective verbs of this type are derived

f rem t he corresponding perfective stems (here hnu-ty) by sitff ixation and eoncornitant vowel

inserriorr. This innovation provides the evidence that both stems were not derived from a

more abstract representat ion of the root morpheme contain ing a f inal {b} ( to be deleted

trr - the Obstnrent f )e let ion ru le in hnu-ty) . See fur ther f r r . 27.ro Simi lar l ] ' in Russian, the basic stem of the corresponding group of verbs contains a

'zero yowel ' (c f . Jakobson 1948). I t is character is t ic that where the \ r*0 a l ternat ion has

beerr e l inr inated, as i1 the CSU verb zua-ty f zu-ut ' (Ol l r r . zz 'ua-t i , f zou'ut t "cal l ' ) , the stem

alternaut rv i th the 'zero vowel ' , i .e. the basic stem, has been general ized.

816 LANGUAGE, \roLuME 4b, NUMBEIT 4 (1969)

(f.) uidi-, zi-, ob'i-, pi.di., roz,i-hraj-ut'; ui.di-, zi-, ob,i-, p,idi,-, rozi-hrtiuaj-ut', uidi-, zi-, obi-pidi-, rozi-hr'ij-ttt'; zi-, obi-znauaj, zi-, ob,i-, roz,i-znaj-ut'; obi-pr'ad-ut'.

Furthermore, there is considerable disagreement between Kyrydenko andAndrusyshen (1957)-the latter dictionary, which is not normative, preferringparagoge in rnany more instances. We may conclude that the synchronic situa-tion is one of a change in progress: stems whieh historically should not haveparagoge are more and more becoming subject to the Paragoge rule.

There is an interesting pattern in the gradual establishment of paragoge irrthese verbs, which rve shall merely note in XII. (a) Wherever paragoge and theabsence of paragoge are both attested in a perfective verb, paragoge is absent inthe corresponding imperfective verb. (b) Wherever paragoge has been establishedin one of the verbs of an imperfective-perfective pair, it is in the perfective verb.(e) Wherever paragoge fluctuates in an imperfeetive verb, it has already beenestablished in the corresponding perfective verb.

XII. (a) ob-kladaj-ut' / ob\i,)-klad-ut',roz-kladaj-ut' f roz(i)-klad-ut', ob-krod,aj-ut' / ob(i)-krad-ut', roz-lcradaj -ut' / roz(i)-krad-ut' .

(b) pi'd,-hrauaj-ut' / pi,di-hraj-ut', roz-hrauaj-ut' / rozi-hraj-ut', roz-znauaj / rozi-znaj-u,t',ob-pr'adaj -ut' / ob'i-pr' ad-ut'.

(c) aid\i)-hrauaj-ut' / uidi-hraj-1d', z(i)Jtrauaj-u,t ' f zi-hraj-utt, ob(i)-hraaaj-ut' f obi-hraj-ut'.

Perhaps there is some coruIection betrveen this primacy of paragoge in the perfec-tive verbs atrd the alternations of paragoge and its absence conditioned by Y-fialternations in the hundreds of prefixed verbs derived from the stems in VI,e.g. z-byuaj-ut' / zi-bj-ut' 'strike off' , z-bEraj-ut' / zi-bra-ty 'gather', z-han'aj-ut' fzi-lma-ty 'round up', z-hynaj-ut' / zi-hnu-ty 'curve', roz-dynaj-ut' / rozi-dlty-tt1'remove the bottom'. But it is difficult to account for in explicit terms.

4.6. LexING. In the Old Russian prefixes {podr} and {ot} (realized as [otr,]before obstruents, nasals, and {l}, cf. $2.2), the fall of the jers left an alternationbetr,veen tense [6] rvhere the final jer had been weak and hence lost, and a lax [o]rvhere the final jer had been strong, and the prefix norv had a paragogic vorvel;e.g. EUkr.p\d-by-tE / podo-b.j-ul, ' 'beatup',6d-by-ty I odo-bj-ut' 'beatback'(ORu.

pod,r-bi,-ti, /podt-buj-iitu, otz'-bi-ti / otn-baj-ilto). Similar alternations 14rere foundin numerous other lexemes in rvhich an original {o} or {e} had in some forms beenfollowed by a weak jer, in others by a strong jer, cf . CSU ueiorok, gen. sg. ue[irlta'evening, dirn.' ; oues, gen. sg. uiusa 'oats' (ORu. ueteroko f ueierzrka, ol)bsb / ouusa).

These alternations were produced bv a mo:rphophonemic rule, the Laxing rule,whieh-as the CSU examples shon'-still exists. Since Earl5'Ukrainian, however,the rule has ceased to apply to the majority of the lexemes that were once sub-ject to it. Thus CSU has, for instance, kinec' f ki.nc'a'end' (ORu. konuca fkonacri), udiuec' f ud'iuc'a 'widower' (ORu. urdouaca f uaclouucci), niltot' f niht'a'nail' (ORu. nogz'tr' / nogrte) rn'ithout laxing. The Laxing rule has ceased to beapplicable at pre{ix boundaries, and as a consequence, CSU {pid}, {vid} arerealized with the reflex of BUkr. {6} also before the paragogic von'el: uid /uidi, pid / pidi.

4.6. TnNSING. Before stems u'hose first syllabic had been a rveak jer, the {o}o f ORu. {c lo } , {ob} , {po} , {p ro} , t1 roz} had become EUkr . {6 } , and the {e } o f

DIACHRONIC \IORPHOPHONEMICS: THE UKRAINIAN PREI-IXF]S 8T7

{bezf , {nr}, {pere} (rvith certain l imitations, cf. $2.6) had become {6}. As a

consequence, E IJkr . {do} , tob} , {po} , {p to } , { roz} , {bez} , {ne} , {pere} were

at f irst realized as [d6], [6b] (or [6], cf. $2.1, $3.1), [p6], [pr6], [16z], [bdz] (and

[r6s], [16], [bds], [b6], cf. $$3.1-2), [n6], [per6] before exactly the same stems beforeg'hicir the Paragoge rule applied-e.g., those listed in VI-VII. These contextual

realizations \vere produced by the Tensing rule, a morphophonemic rule which

also produced alternations in various categories of inflection (cf. noha 'foot',

gen. pl. ni,h; pektty 'bake', m. pret. 7ti,k) and derivation (cf . korouo 'corv', dim.

A'oritko"; do-pek-ty'finish baking', impf. do-7tika.i-ut').We have already seen how the Paragoge rule came to apply to all prefixes

ending in an obstruent, including {ob}, { roz}, and {bez}. In th is rvay {ob},

{roz}. and !bez} have ceased to be subject to the Tensing rule and norv hrlve

I)amgoge instead:

XIII. CSU obi-bra-ty'elect', not +uibraty (ORu. o-bara-ti); rozi-bra-ty ' l ,ake apart', ttot*rizbratu (ORu. roz-bura-ti); obi-Ij-ut' 'w&ter', not *uibl'ul' (ORu. ob-Iai-an); rozi-lj-tr't''porrr out', not *rizl 'ut '(OItu. roz-luj-t itt); obi-spa-ty'sleep in', not *uispaty (ORu. o-sbpa-ti7; rozi-spa-ty(-s'a)'sleep on soundly', not *rispaty (ORu. ro-sz'pa-ti); bezo4n'a'abyss',nol *b i :n 'a (ORtr . bez-dtna).

No te : the ad jec t l e u ib ty j ' r o r rnd ' (d ia l . ) l ? seems to be the on ly examp le wh ich cou ld

show rhat {obf was onee subject to the Tensing ru le, but undoubtedly i t had been lexical-

ized al reat ly in Old I lussian. There are no examples of r iz ' for { rozl , or of b iz- for {bez}.

Irr ri:nyj 'sundr\" (ORti. rozant'), riz- is not prefix but base.

4.?. At sorne time, the Tensirrg rule ceased to be generallv applicable to

prcf ixes. As a con-qequence, {do}, {po}, {p.ol , {ne}, {pere} came to be real ized

as [do]. [po], [pro], [ne], [pere] in all environments:

\ o te : t he rea re nowon l l ' ve r y f ewwordsw i thd i - , p i - , p r i - f o r { do } , { po } , { p ro } ; c f .X IV ' t 8Some of these are undoubtedly lexicalized composita, bul, the verbal formations show that

the 'Iensing rnle still applies to prefixes, although only before a small number of stems-

at least two, at most seven of the 43 listed in VI-VII, according to Kyrydenko.

xI\-. idol: impf . d,o-byraj-ut,, perf. d,o-bra-ty and di-bra-ty / do-ber-zl' 'supplement'

(Olttr. rlo-bara-ti); ilo-han'aj-ut', d,o-h,na-ty (dial. iti-hna-til 'overtake' (ORu. do-gma-ti);

rlo-1ydajttt', di-|d,a-ty 'await' (ORu. d,o-2ada-ti'), do-tykaj-ut', di-tknu-ty and do-tlcnu-ty' totreh' 1ORtr. do-taknu-t i) .

lpol: pi-1rntgky'bl indman's buff ' (ORu. *po-ma\urtky); 'pi- l 'ho 'rel ief ' (ORu' po-luga);

pi-t;ma'darkness' (ORu. po-ttma); pi-hna-ty, see po-hno-ly 'pursue'; pi'hnu-ty, see po-

hntt-t l t 'bencl ' (ORu. po-gz,nu-t i) ; po-rynai-ut ' , pi-rnu-ty'dive' (ORu' *po-ranu-t i , cf ' po'

r inu-t i) ; pi ,-sla-t11, see po-slo-fy 'send' (ORu. po-srla-t i) ; pi-tknu-ty(-s'a), see po-tknu-ty

( -s 'a) 's tumble ' .

iprcri : pr i-zu-ys'ko, pri-zu-ySde 'surname' (ORu. 'pro-ztu-i \ te); pri ' rua'precipice' (ORu.

p r 0 - r f l a ) .

In add i t iou to the words l i s ted in X I \ ' , we must ment ion d , i -zna j -u t ' ' exper ience ' , d i -

i; llrindenko (1909, s.v.) cites the word from Hucul sources and glosses it 'made from a

whole log'. Cf. ORu. obtl t ' round' ( *ob-utl t (cf. . \rasmer 1953-57, s'v. oblyj) 'rs Irr \\-I we reproduce Kyrydenko's treatment of doublets. Where d'i- and do- co-exist,

the dict iolary gives both forms separated by'and'; their order presumably ref lects the

etlitor's prefererrce. Where pi- and.po- co-exist, the form with pi- is entered separately with

a cross-reference to the maiir entry for the verb, but is not cited there. It is not clear whether

this r l i f ferent treetment of di- andpi- ref lects l inguist ic real i ty or merely a change in edi-

t o r i a l p rac t i ce .

818 LANGUAC}II, VOLUME 45, NUl,{BIrtit, 4 (1969)

stan-utt 'obtain' , p' i -znaj-ut"recognize' (with derivatives), in which di-, pi- occur befgrestems that begin with an original consonarrt eluster. These examples are easy to untlerstandonce one remembers that the Tensing rule originally applied in the same environments asthe Paragoge rule. Just as the Paragoge rule became applicable before stems beginning withan original consonant cluster, so did the Tensing rule. I t is quite possible that {dol, {pol,lproI once were realized as [d6], [p6], [pr6] also before other stems with an inherited initialconsorrant cluster, but have changed baek to [do], [po], [pro] as a result of the general cur-tailment of the Tensing rule.

CSU has no archaisms with n' i- or peri- for {ne}, {perel. nAfor Ine} is occasional l .y at-tested in tr)arly Ukrainian texts, e.g. da n'A jznes'Al' s'o 'may it not be removed' (ORu. done jaz-neselr sri) on Euphrosynia's cross, 1lO1 (cf. Lehr-Splawiflski lg65:80).

4.8. PenAGocIC {i}. We saw above horv the domain of the Paragoge rule haschanged since thc fall of the jers, so that it now applies in environments u'hichitr some respects are narrorver, in others wider, then originally (cf. $$a.1-4). trVernust now consider the question of horv the rule was modified so as to assign aparagogic {6} (corresponding to CSU {i}) instead of the original EUkr. {o}-the regular reflex of a strong jer.

The original paragogic {o} is still preserved in a number of lexical items, insome as the only possibilitv (XV), in others as a less usual or a dialectal variantbeside {i } (XVI).

XY. bezo-da'o 'abyss' (ORu. bez-d,tna), airJo-zua'appeal' (ORu. otz,-zz,ua), uid,o-msty-tu'avenge' (ORu. ata-must i ' - t ' i ) , zo-un' i ' f rom outs ide' (ORu. jw utnd), zo-as ' int . 'qui te ' (OIt1.sz' uasdnrc), zo-ml'i j-ut' ' taint' (ORu. sz-moldj-uta), zo-tl ' i . j-zrl ' 'smolder' (OR1. s*tuld-r.ita),p ido- lda-ty 'awai t ' (ORu. poduirda-t i . ) , p i .do-zry- ty 'suspect ' (ORu. pod.u-zard-t i1, p ido-sp'i j-y1r 'be in time' (ORu. podr-spdj-titt), pido-5ao ,sole, (ORu. pod,z.-Eaua).

XVI. zo-ujanu-ty 'wi ther ' (OItu. st -ur inu-t i ) , zo-hna-ty ' rourrd up' (Ol l r - . sa-gtn.a- t , i ) ,zo-hr'i j-ut' 'warm' (oRu. stgrdj-rita), zo-star.t1-ty 'age' (oRu. sz,-stari-ti), zo-tnu-1,t1 ,cut

down' (ORu. sz'-tun-utu), obo-ml'i j-ut' ' faint' (ORu. o-mold.j-rita), pido-bhctj-ut' ' fold uncler'(ORu. podt-gtbaj-titr), pido-hrij-ttt ' 'wa,rm up' (ORu. poclt-grdj-titt), pitlo-sla-ty 'send rrp'(ORu. pod,o-sz,la-ti), rozo-rnl'. i j -ut' .weakerl' (ORu. roz-mtli j -zita).

To the latter category one mnst add the fixed prepositional phrases in which paragggicfo| may be preserved before the pronominal stems mn-oju 'me ( instr . sg.) 'and ues' ,a, l l ' :

zo (nado, pido) mnoju (oRu. st mtnojti), zo (na.d,o, pido) us'im, tts'i jeju, as'ima, bezo (zo)u8'oho, usej i , us ' ' i r (e.g. ORu. st ar ,sdma,bez atsego). Simi lar l l 'before us 'akyj 'any k i l t l of ' :nad,o us'akym, bezo us'akoho (ORu. nadz, uusticl,mu, bez ar,stikogo).

4.9. The change from paragogic {o } to {6 } is a simple one from a purelydescriptive point of view; it can be described as the replacement of one dis-tinctive feature value ([*lax1; by its opposite ([-lax]). But there is no re&so1to assume that morphophonemic rules are changed simply by the arbitraryreplacement of one distinctive feature value by another in some part of the rule.

The development of {obl, \rca}, {bez} permits us to make an important in-ference which can help us to explain the change of paragogic {o} to {d}. Theabsence of anv residual forms u'ith *uib-, *r,iz-, or *biz- (let alone *uibi-, *riei-,*bi,zi'-) indicates that { ob } , { rozl , and {bez } became subject to Paragoge insteaclof Tensing, not in addition to Tensing (cf. $$4.2, 4.6). This suggests that theParagoge rule preceded the Tensing rule in Early Ukrainian, and not vice versa.

After {ob}, {roz}, and {bez} had ceased to be subject to the Tensing rule,t h i s r u l e a p p l i e d o n l y t o t h e f i n a l { o } a n d { e } o f { d o } , { p o } , { p . o } , { r r . } , { p e r e } .

DIACHRONIC MORPHOPHONEMICS: TIIE UKRAINIAN PREFIXES 819

Since it follou'ed thc l'aragoge rule, the Tensing rule had to refer to the absenceof a rnorpheme boundary before the {o} of {do}, [po], {p.o} in order not toapply to the paragogic {o}. A simplification of the ruie u.ould rnake it applicableto any prefix-final {o}, including the paragogic {o}. This seerns to be nhathappened.le

Once the Tensing mle applied to paragogic {o}, howevcr, the pa.ragogic {o}could not continue to exist as {o}, for succeeding generations rr-ouid sirnplyinterpret it as a paragogic {6}. This explains u.hy, rvhen the Tensing ruie ceasedto be applied, the Paragoge rule continued to assign a paragogic {6 }.

4.10. Suuueny. We can no\^,' sum up the various changes discussed in thepreceding pages and try to define their chronological rclations. to the extentthat these can be established.

a. The Obstruent Deletion rule has ceased to be generally applieable. It norvapplies on-ly to {ob} and only before certain stems (cf. $3.1).

b. The Tense-before-Tense rule has been dropped (cf. $3.2).c. The Paragoge rule has become appl icable af ter lu l , {bez}, { roz}, {ob},

i.e. after all prefixes ending in an obstruent (cf. $a.2).d. The Paragoge ruie has become applicable before a number of stems begin-

ning ri-ith a conson&nt cluster rvhich did not originally contain a weak jer (cf.

$1 .4) .e. The Laxing rule has ceased to apply to prefixes (cf. $4.5).f. The Tensing rule has become iimited to {do}, {po}, {p.o} as a collsequence

of Change c (cf . $$4.2, 1.6).g. The Tensing rule has become applicable before stems beginning with con-

sonant clusters rvhich did not originally contain a weak jer (cf. $4.7).h. The Tensing rule has become applicable to paragogic {o} rvith the result

that the latter changed to {6} (cf. $4.8).i. The Tensing rule has ceased to be generally applicable and norv applies

onlv before a ferv stems (cf. $4.7).j. Finally, the Paragoge rule has ceased to apply to prepositions except before

trr-o or three stems (cf. the examples in $a.8).20 The preposition \zl does, horv-cver: receive paragoge before sterns beginning u'ith sibilant * consonant. Thispart of the Paragoge rule does not seem to have changed since the fall of thejers (cf. $4.3), except for the change to parzr,gogic t6), nou' {i}.

4.LL. cHRoNoLoGy. At several points during the preceding discussion vvementioned thc cxistence of residual forms, lexemes, or fixed. phrases u'hich failedto unclergo one change or another. These, in addition to the internal relationsrlmong the changes, permit us to establish a partial relative chronology for thechanges.

The extension of Tensing to the paragogic {o} before all relevant stems-this

le In S4.1 I s tated that the paragogic vowel might be descr ibable eibher as paragogic

or as prothetic. It is the fact that it became subject to the Tensing rule that permits us to

infer that i t was anal l 'sed as pref ix- f inal , i .e. paragogic.20 I t is possib le, even l ikely, that the Tensing ru le was s imi lar ly curta i led before Change

( i ) occ , . r r red . Bu t th i s ca r r r l eve r be known.

820 LANGLTAGE, I/OLUMIii 45, NUI\IBI]I1 4. (i9i'1J)

is Change h-rn';st linve prececled the severe curtailment of thc Teusirig rule(Change i).2'

Archaisms rvith paragogic {o} shorv tirat the tivo cxten-*ions of Paragoge pre-

ceded the change to paragogic {6}.I''orms like zo-u'n'i 'from rvithout', bezo-tln'a'al:yss', 'rozo-ntl'ij-al' 'rveaken', obo-ntl'i,.j-ul,' 'faint' shorv that Change c preceded

Change h. pir!,o-spi,j-ut' 'be in time', pido-ltr''ij-'ut' 'warmup', zo-lr''ijut' 'warm',

zo-ujanu-tg 'rvither' shorv that Change d also preceded Change h. Undoubtedlythe extension of the Tensing rule to stems beginning rvith inherited consonantclusters (Change g) rvas contemporary with the parallel ertension of the Paragogerule (Change d).

Synchronicaily, ttre Obstruent Deletion rule precedes the Pa,ragoge rule-hence all occurrences of {ob} rvith the {b} intact receive paregoge before therelevant stems. But this fact does not justify any inferences concerning therelative chronology of the curtailment of the Obstruent I)eletion rule (Change a)and the extension of Paragoge to {ob} (Change c).

The loss of the Tense-Before-Tense rule and of the Laxing rule has left noarchaisms. Perhaps these changes occurred eariier than the ones discussed above.If this is the case, we c&n define four chronological phases in the history of themorphophonemic rules applicable to prefixes, each phase characterized by oneor more morphophonemic changes, as belon'.

XVII. Phase I: Changes b, e.Phase I I : Changes c (and f ) , d, g.

Phase I I I : Change h.Phase IV: Change i.

In this w&y, an attentive analysis of the linguistic data yields a clearer picture

of the chronology of the changes than could a study of the textual attestation ofthe prefixes. Due to the peculiarities of the {Jkrainian spelling system prior tothe nineteenth century, the termini ante quos provided b-v ehanges in the spellingof prefixes are mostly so late as to be rvorthless.

5. In the preceding pages we have sketched the historical development of asection of the morphophonemic component of lJkrainian, consisting of ten de-finable changes which have occurred since Early lJl<rainian, i.e. over the last800 years. As rve have seen, most of these changes have occurred in a chrono-logical order which can be established with certainty. Anv order other than theone outlined in $1.11 u'ouid have yielded different realizatiotts of the prefixes

irr CSU-in particular, rvould necessarily have produced different kinds of

2r In $1.3 we excluded from our discussion the Old Russian prefix aa, which merged withORu z. In a small number of prefixal formations, CSU {u} has an obsolescent alternant,

uui-, apparently a blend of the two preconsonantal realizations we would assume for EUkr.

{vf , [vo] with paragoge and [ti] without. Kyrydenko recognizes uui- with eleven of the stems

I isted i r r VI and VI I , e.g. uui-bhaj-ut ' 'bend' (ORu. uz ' -gtbaj- t i ta) , uui- rua-ty(s 'a) 'break'

(ORu. az-rzua-t' i). But in all but one of these-uai-ssa-ty 'take in' (ORu. ua-sasa-ff)_-he

prefers the alternant ru-. It is a- that occurs with the rest of the stems in \rI and VII. It is

also replacinguui in fixed prepositional phrases likeuui, sn'i ' in a dream' (ORu. ua stni.),

uui l 'uoai ' in Lvov' (ORu. uz, Lwoud). From the point of v iew of chronology, i t is in lerest ing

to note that CSU has no archaisms wi th the or ig inal paragogic foI preserved.

DIACHRONIC NIORPHOPHONEI\'IICS: THE UKRAINIAN PRIIITIXES 821

residual forms than the ones we have accounted for in the course of the above

discussion. We can be confident, then, that the changes u'e have reconstructed

took place as described.5.1. In $$4.10-11 rve considered these changes as if each of them were a dis-

crete event, taking place between trvo successive states of the language. We

knorv, holvever, that this is not the rvay morphophonemic changes talie place.

\Iorphophonemic change presents the same dichotomy as all other kinds of

iinguistic change. It appears to be abrupt-a (mutation', as Jakobson would

say (1962:205, 333, 652)-if rve limit ourselves to a comparison of successive

states of the morphophonemic system; a given state either has or does not have

a certain morphophonemic rule. But in the messages produced brr successivegenerations of speakers changes appear gradual. This we knovt' from the history

of languages: rvherever rn'e have continuous textual attestation, morphopho-

nemic change is attested as gradual. We also know this from dialectology: mor-

phophonemic changes in progress are reflected as spatial gradations. And we

knorv it from synchronic descriptions of languages: in different styles, an ex-

isting morphophonemic rule will often be applied rvith different degrees of con-

sistencl'. Indeed, several of the changes discussed in the preceding pages have

l,ielded a sl.nchronic situation rvhich is incomprehensible except as an inter-

mediate stage of :r gradual change in progress.

It is clear that arl account of morphophonemic change cannot be considered

adequate if it merelv defines differences between successive states of a morpho-

phonernic s)'stem. It must also shorv horv these differences have come about.

In the follou'ing pages u-e ri-ill consider the changes u'hich rve have defined ex-

plicitlv as processes, and u-ill tr1' to determine rvhat were-and are-the mech-

anisms of change.5.2. T1'pologicalll' the changes rve have discussed fall into three categories.

1. The domain of a rule has been extended. We have seen four examples of

th is tvpe:

1a. The Paragoge rule has become applicable before some stems beginning with inherited

corrsonant c lusters (cf . $a.a) .1b. The Tensing rule at one time was extended in exactly parallel fashion (cf. $a.7).

lc. The Paragoge rule has become applicable after all prefixes ending in an obstruent

(c f . $a .2 ) .

ld. Il it ially applicable only to the {o} of prefixes, the Tensing rule was extended to apply

a lso to pa ragog ic {o } ( c f . $4 .8 ) .

2. The domain of a rule has been curtailed. We have seen the following ex-

amples of this:2a. The Tense-before-Tense rule has been dropped (cf. $3.2)'2b. Brrth the Paragoge rule and-before Change 2c occurred-the Tensing rule have

ceasecl to be applicable to prepositions in general. The Paragoge rule continues to apply

af ter pref ises (cf . $-1.10 arrd fn. 20).

2c. Tire Tensipg rule has ceased to be applicable before all but a handful of stems (cf.

$.1.7) , a l though i t cont inues to be appl icable in inf lect ion and in suf f ixal der ivat ion '

2d. The Laxing rr.i le has ceased to apply to prefixes (cf . $4,5), although it, continues to be

relevant in inf lect ion and der ivat ion.

2e. The Obstnrent l )e let ion ru le has ceased to be appl icahr le to pref ixes other than {ob},

ancl i t appl ies onlJ ' before certa in stems (cf . $3.1) .

822 LANGUAGE, !'OLUMtr 45, NUMBER 4 (1969)

3. The effect of a ruie--what feature(s) it assigns-has changed. We have seenonly one example of this, the change of paragogic {o} to {6} (cf. $4.8).

5.3. Changes 1a and lb as listed in $5.2-the extension of Paragoge andTensing before stems beginning rvith inherited consonant clusters-is our clearestexample of hoiv the domain of a rule is extended. The environmerrt definition ofthese two rules, as it refers to the stems before rvhich they apply, must have hadthe very general form it nolv has ever since these rules were estabiished at thefall of the jers. The extension of the domain of the rules has been a lexeme-by-lexeme extension-as far as the Paragoge rule is concerned, apparentll' still inprogress-to stems which already conform to the environment definition of therules ('beginning rvith a consonant cluster'), but rvhich for historical reasonsconstitute exceptions to its application.

Change la shorvs what happens when a morphophonemic rule defines itsenvironment in too general terms. When this is the case, lexical items which forhistorical reasons are not subject to the rule (i.e. do not constitute an environ-ment for its application) have to be specified as exceptions to the rule. Theextension of the domain of the rule consists in gradual elimination of such ex-ceptions.22

Change la, then, permits us to identify tu'o phases in this type of morpho-phonemic change, an over-generalization in the formulation of the environmentdefinition of a rule, followed by the elimination of the lexical exceptions neces-sitated by this over-generalization. The first phase can be understood as moti-vated by the need to formulate grammatical rules in as simple (general) termsas possible. The second phase follorvs as the traditional forms are imperfectlypassed on to succeeding generations of learners. It entails a simplification of thelexical component of the grammar. The first phase is necessarily abrupt-butit is covert and goes unnoticed by the speakers. The second phase is gradual andmay be more or less apparent to the speakers as older and newer forms areutilized as indices of style (cf. $5.1).

Changes lc and ld differ from Change 1a only in that they are no longerchanges in progress; they have been completed. Both of the rules involved inthese changes were established at the fall of the jers and at that time had toaccount for the irregularities of the inherited paragoge and tensing respectively.The Paragoge rule--to use Change lc as an example-could have been formu-lated in such a way that it applied only after prefixes specially marked as enyiron-ments for the rule (ORu. {nadr}, {ot}, {podr}, {s"}). But from the historical

2r Put in very concrete terms, the process can be envisaged like this. After the fall ofthe jers, every learner has to formulate a rule to govern paragoge. He easily defines theonly shared property of the stems before which paragoge is found, viz, ant initial consonantcluster. When there is no paragoge before a certain stem beginning with a consonant cluster,he views that stem as an exception. Since the inherited set of exceptions cannot be produced

by the rules of the grammar, they must be learned individually. Whenever & speaker hasnot learned a certain stem as an exception, he can only produce its respective forms ac-cording to the rules. Whenever the speakers fail to correct a learner who produces formsaccording to the rules instead of using the traditional forms, the perpetuation of the in-her i ted set of except ions is in jeopardy.

DIACI{RONIC },IORPHOPHONtrMICS: THE UKR,AINIAN PRtrFIXES 823

deveiopment it is clear that, as long as the rule has existed, it has defined its

environment in the general terms ('after prefixes ending in a consonant') in

rvhich it is defined today. The prefixes after rvhich the Paragoge rule historically

should not apply (ORu. {jur}, {bez}, {roz}, [ob]) must then originally have

been marked as exceptions, and the extension of paragoge after these prefixes

consisted in the gradual omission of these markings. In short, Changes lc and

1d illustrate exactly the same mechanism as Changes la and lb.

5.4. The curtailment changes are in all essential respects exact counterparts

of the extension changes, the only important difference being the directionalit5'

of change.Let us consider first the elimination of the Tense-before-Tense rule, Change

2a. If we suppose that this rule in its original formulation applied only to mor-

phemes specifically marked as subject to the rule, then we can understand its

curtailment-and eventual elimination-as the result of a simplification of the

lexicon, exactly analogous to the one we defined for the extension changes. Here

then, as in the extension changes, we have trvo phases: first a morphophonemic

rule is formulated in such a way that individual lexical items have to be marked

with respect to that rule; then the marking of these items is gradually omitted.

The difference is that where the marked items are exempt from the rule, the

simplification of the lexicon results in an extension of the domain of the rule;

but.w'here the marked items are the ones that are subject to the rule, the result

of a simplification is a curtailment of the domain of the rule.

In $5.3 we saw that an extension change starts with an over-generalization in

the formulation of the environment definition of a morphophonemic rule. It is

natural to assume that curtailment changes correspondingiy begin with under-

generalizations.In the elimination of the Tense-Before-Tense rule, then, we have the special

case of under-generalization where there is no generalization: the rule defines

no shared property of the morphemes that constitute the environments in which

it is applicable (beyond their having a final obstruent), but applies exclusively to

marked items. Changes 2b-e, on the other hand, involve morphophonemic rttles

rvhich originally applied both in environments defined in general, but too narrow,

terms and-to compensate for the under-generalization-to certain specially

marked items.Change 2b can serve as an illustration. At the time when the Paragoge and

Tensing rules were established, they could have been formulated to apply both

to prefixes and prepositions alike before any morpheme beginning with a con-

sonant cluster. But apparently they rvere not formulated in such general terms.

It was only with prefixes sensu stricto that paragoge became established before

morphemes defined irr general terms. Prepositions, on the other hand, were

subject to paragoge only before certain marked morphemes. While paragoge

continues to be assigned in the environment defined in general terms (i.e. after

prefixes sensu stricto before stems beginning with a consonant cluster), it has

ceased to be assigned after prepositions before the vast majority of the mor-

phemes that must once have been marked as constituting an environment for

824 LANGUAGII, VOLUI4E 45, NUMBtrR 4 (1969)

paragoge (e.S. aU substantives originally having a rveak jer as the first syllabic).In CSU, prepositions receive paragoge only before three or four specially markedmorphemes (cf. $4.8 and fn. 21).

5.6. Extension changes a,nd curtailment changes then can be understood asmanifestations of one and the sarne mechanism of change. They differ essentialiyonly in that extension changcs spring from over-generalizations in definitions ofenvironment, curtailment cha,nges from under-generalizations. Since both thesetypes of change result in changes in the relative frequency of morphophonemicalternants, \\,c mav call them euANTrrArivo types of change-in contradis-tinction to the last type rve have to consider, euAlrrArrvE change.

In a qualitative change, the effect of a morphophoiremic rule, i.e. the feature(s)it assigns, is changed. While quantitative change involves the relation betweenthe environment definition of a rule and the lexicon, qualitative change arises asa consequence of interferencc betiveen morphophonemic rules: the first of trvorules changes so that it comes to assign directiy l'hat rvas formerly the cornbinedeffect of both of the rules. The change may imply the elimination of the secondof the rules; this is the case rvhen the second rule applies only in environmentssubsumed in the environment definition of the first rule. In such cases, the secondruie is, as it were, incorporated into the first. If, horvever, it is applicable toenvironments other than the one(s) covered by the first ruie, the second rulelvill remain in the grammar after the change-as in our example, the change ofparagogic {o} to {6} (cf . $4.8).

5.6. These, then, are the three basic types of morphophonemic change. Amorphophonemic rule preserrts onlv two possibilities for change. There can bea change in the features assigned by the rule: this is qualitative change. And therule can become applicable in a different domain: this is quantitative change.In the latter instance, there are tu'o possibilities: the domain of the rule may beextended or curtailed.2s

6. The results reached in the preceding sections-the reconstruction of thehistory of the Ukrainian prefixes, the typology of morphophonemic change, andthe account of the mechanisms of change-have significant implications. Wervill not discuss here the specific implications for the history of Ukrainian, butrvill limit ourselves to two points of general interest.

6.1. In the first place, u'e have seen that extension and curtailment changesinvolve the same mechanism of change, and can be understood as identicallymotivated (both entail simplifications of the lexicon) if we account for theirdifferent directionality by recognizing tu'o possible relations betrveen rule and'exceptions'. Individual morphemes may be exempt from a rule although theysatisfy its environment definition. When this is the case, the domain of the rulewill be extended as thesc morphemes lose their exceptional status. If, on theother hand, individual morphemes are specified as subject to a rule although

23 To these three basic types of morphophonemic change we may add the possibil ity thattwo or more rules may be re-ordered with respect to one another (with or without any changein their formulation). The example discussed in $4.9 might superficially look like an exampleof this type of change. However, there is no evidence that the order of the Tensing and Para-goge rules was ever different from the one attested by the development of {ob} , lrozl,{bez} and the change of paragogic {o} to {6} .

DIACHROI\IC IIORPIIOPTTONEI{ICS: THL UKITAINIAN PIIEIiIXES 825

the-v do not satisfy its gerrerai environment defiriition, the clomain of the rule

will be curtailed as these morphemes lose their exceptional status.In order to account for the fact that individual morphemes ma)r be exceptional

in these two rvays, one rnay assume that arly morpheme in a language is specified,rvith respect to every rule in the grammar, a,s subject or trot subject to that rule.

Chomsky & Halle (1968 iI72 tr.) formulate a convention to this effect and furtherassume that it is regular for all morphemes to be subject to all rules. In terms of

the theory of markedness, in their rvords, '[f rule m] is the "unmarked" valuefor the feature [rule rn,], for each m' (374, fn.) That is, morphemes that are sub-ject to a certain rule are unmarked li'ith respect to that rule, rlhereas morphemesthat are not subject to it are marked.

If we accept this assumption, we can understand the second phase of extension

changes, sketched above, as a reduction in the number of marked morphemes-

i.e. as a simplification of the grammar. But curtailment poses a problem. If we

accept Chomsky & Halle's assumption, we have to vierv the morpheme-by-

morpheme curtailment of the domain of a rule as a gradual accumulation of

exceptions to the rule, which gradually adds to the complexity of the grammar

until the rule can be reformulated. Vierved in this way, a curtailment change

may rvell lead to a simplification of the grammar-namely, when the original

rule is reformulated; but this appears a paradoxical effect of the progressive

complication leading up to it. Even more seriously, if curtailment is conceived

as an accumulation of exceptions, it is impossible to understand how any

morpheme-by-morpheme curtailment could ever be initiated or proceed-for

in terms of this concept, each new exception in a learner's grammar would be

motivated neither by the corpus he had analysed (from which it would deviate)

nor by the rule he had himself formulated (to rvhich it would be an exception).

It would seem that, in formulating a grammar, a learner cannot have any moti-

vation for marking morphemes as exceptions to his rules other than the necessity

to produce traditional forms which his rules rvould otherwise not generate. The

idea that a learner would spontaneously create exceptions to his own rules, as

entailed by Chomsky & Halle's assumption, is unacceptable.Let us assume instead, in accordance u'ith our analysis of the two types of

quantitative change, that in a synchronic grammar a morphophonemic rule

may define its environment either as an over-generalization or as an uncler-

gerreralization. It is natural to assume that, in either case, rnorphemes rvhich

satisfy the rule's environment definition and are subject to it are unmarked with

respect to that rule. \{orphemes which are exempt from a rule aithough thev

satisfy its environment definition, or rvhich undergo a rule aithough they do not

conform to its environment definition (or r,vhich undergo a rule that has no

general environment definition-cf. $5.4), will be marked u'ith respect to the

nile in question.2af,et us also assume, finall-v, that each morphophonemic rule specifies'ivhether

2a Morphophonemic rules which have no general definition of environment, but apply

only to marked items, undoubtedly are particularly important during the acquisition of

language. It is reasonable to assume that all morphophonemic rules are first formulated in

these terms, and are only gradually-as the learner acquires enough material to justify

generalizations-reformulated to refer to the shared properties of the provisionally marked

items. Some rules mav never be so reformrilated.

826 LANGUAGI!, \'OLUIIE 4b, NUN{BI'II 4 (1969)

marked items ere exempt from the rule or subject to it.25 By this assumption, itis a property of each morphophonemic rule rvhether its domain will be extendedor curtailed, and it becomes possible to interpret extension change and curtail-ment change as identical in nature: in both types of change, the number ofmorphemes marked rvith respect to a certain rule is reduced.

Although prompted by an analysis of change, this last assumption is equallymotivated by the facts of synchrony. It is tantamount to including in everymorphophonemie rule a statement of its productivity; and only if we assumethis information to be included in every morphophonemic rule can we accountfor the way in rvhich a native speaker handles morphemes that are new to him,neologisms or not. It is a, fact that such morphemes, which for obvious reasonscannot be marked, are not subject to all the rules (as Chomsky & Halle's theorywould have it), but undergo some rules and are exempt from others. This canonly be because the native speaker kno$'s, for each rule in his grammar, whetherit is or is not to be applied to unmarked items.

Returning to the question of morphophonemic change, we have seen that ourdistinction between productive rules (rvhich affect, unmarked morphemes) andunproductive rules (which only affect marked items) makes it possible to de-scribe the morpheme-by-morpheme curtailment of the domain of a morpho-phonemic rule, from the very first deviation from the received pattern, as asimplification. This implies that, before this first deviation is established, therule must already have ceased to be productive for the environment in question.The question of holv morphophonemic rules can thus become unproductivecannot be discussed here in detail, for it is inseparable from the larger problemof the relation betrveen phonological and morphophonemic rules. But perhapswe can outline an ans\,yer to this question.

It is characteristic of phonological rules (as distinct from morphophonemicrules) that they normally apply without restriction to all words in a language.That is, phonological rules are productive in the sense just described. This isclearly shown by the usual treatment of foreign words. Foreign morphemes maybe marked as exempt from certain phonological rules; but as they are assimilatedand lose their exceptional status, they become subject to all phonological rules.By contrast, curtailment and uitimate elimination is the most usual fate ofmorphophonemic rules. Indeed, this is so much the normal development thatwe cannot take the productivity of morphophonemic rules for granted. We mustinstead ask why morphophonemic rules normally are unproductive, but ex-ceptionaliy may retain a measure of productivity.

One may assume that phonologicai rules, when they are introduced into agrammar, manifest reiations in the hierarchy of distinctive features of the lan-guage at that stage, and that it is their function u'ithin the phonological systemto produce overt signs (allophonic variation, neutralization) of the distinctive

26 It should be noted that, where a rule applies in several different environments, thetreatmentof marked morphemes may have to be specified separately for each environmenb.The development of the Paragoge rule il lustrates the relative independence of differentparts of a complex environment definition as regards the direction of change; cf. $5.2.

DIACHRONIC N{ORPFIOPHONEMICS: TITII UKRAINIAN PITEFIXITS 827

feature relations that define the phonemes.z0 As long as the underl;ring relationspersist r,l'hich motivate a given phonological rule in this \vily, the ruie remainsproductirre. But if these rclations change, the rule loses its sign function andhence its phonological raison d'6tre. It, or rather a homologous rule, may stiilhave to be set up to produce the inherited distributional regularities; but theformulation of this (morphophonemic) rule u.ill be determirred not by relationsthat are intrinsic to the phonological system, but by the occurrence of specificalternations in the corpus on the basis of u.hich the rule is forrnulated. Whenthe existence of a rule is thus predicated on the occurrence of specific rnorphemesin the corpus, rather than on the constitutive relations of the phonologiealsystem, it has ceased to be applicable r,vithout restriction to all words in thelanguage; that is, it has lost the productivity characteristic of phonological rules.The assurnption that phonological rules have a sign function thus accounts bothfor the productivity of phonological rules and for the loss of productivity u'hena phonological rule is reformulated as a morphophonemic rule. The questionunder what circumstances morphophonemic rules may nevertheless be pro-ductive must be unanswered here.

6.2. One of the incidental developments in generative phonology has beenthe discovery that the phonetic output of a grammar can be derived from ahighly abstract representation, rvhich in some respects may resemble a muchearlier stage of a language. As examples, in addition to Chomsky & Halle 1968,one can mention Halle 1963 and Lightner 1965, ivhich deal u'ith Russian. Asthis work shows, it is possible to postulate for Russian a 'systematic phonemic'representation rvhich strongly resembles the eariiest stage of Common Slavicrecoverable by internal and comparative reconstruction.

There is no reason to believe that a similar system of notation could not beused for CSU, given an appropriate set of rewrite rules and a judicious use ofdiacritic features. But as a hypothesis about the linguistic competence of speakersof Ukrainian, such a 'phonology'u.ould not be valid. This we can conclude fromthe changes we have anaiysed..

One of the characteristies of a 'systematic phonemic' representation of CSUwould be the use of a pair of segments analogous to the short *i and *u of Com-mon Slavic, the jers of Old Russian. These segments would have to be used inthe first syilable of all the stems listed in VI and VII, and in none of the stemslisted in X. It is not difficult even for a moderatelv skilled linguist to separatethese two categories of stems by internal reconstruction. But the fact thatparagoge has been extended to the stems in X (cf. $a.a) is an unmistakable indi-cation that the grammars of native speakers of Ukrainian do not distinguishoriginal initial consonant clusters from consonarlt clusters in which a jer hasbeen lost.

Similar arguments can be made on the basis of some of the other changes wehave examined. The vocalism of {pid}, {vid} as compared rvith {ob}, lrozlshows the analvst rvhich prefixes ended in a jer and u'hich did not. Still, in Early

2s This assttmption and some of its most important implications are discussed in Andersen1969d. It serves as the theoretical forrndatiorr for the method of analvsis used irr Andersen1969b . c .

s28 LANGUAGIi, VOLUI\{U 45, NUN{Brln 4 (196-0)

Ulirainian, paragoge was cxtcrrcletl to therrr all. Again, the 'motrilit','' oi thcparagogic lol of Eari5, l l l irainiari and the'stabil ity'of the vorvel of lclui, lpof ,lprol rvould indicate to the linguist that Tensing and Parngoge should be appiieclin this order. Still, the paragogic {o} became subject to Tensing.

The high degree of abstraction characteristic of the 'systematic plionemic'

representation used b1, Halle and Lightner is eviderrtly motivated bl' a desireto 'miss no generalizations', i.e. to account in the rules of the 'phonologl" fornll observed distributional rcgularities. This is laudable, but lve must recognizethat, at any given stage, not ail distributional regularities are accessible to theiearner. If they \vere, each historical change u'ould have to be conceiverl of as arviilful distortion of the inherited pattern, rvhich rvould be absurd.

Our study of the lJkrainian prefixes proves-as any investigation of morpho-phonemic change u'ould-that in formulating morphophonemic rules, the learner

fails to exploit some of the more abstract regularities observable in the corpus he

is analysing. The rules he formulates refer, instead, to regularities closer to thephonetic surface (stem-initial consonant clusters, or prefix-final consonants, in

the case of the Paragoge rule-cf. $$4.2, 4.4); or do not 'capture any generaliza-

tions', but refer to lexical markings. E.g., {6d} was once marked as subject to theTense-before-Tense rule, but {pdd} and {nad} were not, cf. $3.3; {ob} undergoes

Obstruent Deletion before marked stems only, cf. $3.1.2?What does this sho'w? It shou's that, in actual grammars, morphophonemic

alternations are not produced from a very abstract 'systematic phonemic' repre-

sentation by a set of rules which codify all generalizations that the output of thegrammar might justify. Rather, it seems, the rules that generate morphopho-

nemic alternations operate on a phonological representation of a much lower

degree of abstraction, not very different, perhaps, from that of the 'morphonemes'

of Halle 1959. If this is so, it is important evidence for the existence of a significant

level of phonological representation on a lower degree of abstraction than 'sys-

tematic phonemics'-not intermediate betweer] a 'systematic phonemic' repre-

sentation and a phonetic notation in the sense that it would be derived from the

former, but of primary importance as input both for the morphophonemic

component and for the phonological rules.6.3. In conclusion, let us return to our deseription of the mechanism clf mor-

phophonemic change in $5.3, for it is on the background of this view of morpho-

phonemie change that the remarks in the last few pages should be understood.

In $5.3 we identified two phases in each morphophonemic change: a covert

phase, consisting in the formulation of a new morphophonemie rule, atrd att

overt phase, consisting in a gradual elimination of lexical exceptions to that rule.

The distinction betrveen these two phases is of fundamental importance, for it

is relevant for all linguistic change.

rz The change of EUkr. -hybaj-ut' to CSU -hynaj-ut', which I mentioned in fn. 15, isrelevant here too. fn terms of 'systematic phonemics', the relation between llUkr' -hnu'-t11

and -hybaj-zt'would be considered regular. The two forms would be derived from (gub*

nou*ti) and ((9t7bf 6i*e)*n*ti)-or the like-by what would be considered perfectly

general rules, and -hybaj-al' would be derived from -hnu-tA by lengthening of the root

vowel, a regular process in imperfectivization. The fact that -hybai'ut' was ehauged to

-h,gnaj-ut', however, shows us that the speakers of this language did not make use of these

supposed regularities, but of some very different ones (cf. fn' 5).

DIACHRONIC MORPHOP}IONEMICS: THtr UKRAINIAN PRtrFIXES 829

The first (covert) phase we rnay call rxoucrrvn cHANGE, for it arises out of theinductive process of rule formulation. This is necessarily abrupt, for a ruie iseither identical to or different frorn a corresponding rule in the grammar fromwhose output it is induced. The second phase \ve may call opoucrrvn cHANGE,for it takes place in the process of creating surface forms from base forrns by theapplication of rules. Deductive change is overt, for the gradual eliminatiorr oflexical exceptions is reflected in the speech of successive generations of speakers.In inductive change, a rule is reformulated without any change in the output.Deductive change, on the other hand, modifies the output so as to make it corre-sporrd more closely to the rules of the grammar.2s

These two different aspects of linguistic change present the investigator withdistinct problems. In morphophonemics, deductive change-that is, change inthe output of the grammar-can be explained only on the basis of the rules ofthe grammar that produces the nerv forms. In $6.1 we saw horv the changesanah'sed in this investigation shed light on the process of deductive change,permitting us to understand horv the direction of morphophonemic change-thequestion of rvhether the domain of a rule will be extended or curtailed-is deter-mined by the productive or unproductive character of that rule in the synchronicgrammar. On the other harrd, u'hen an inductive change occurs, it can be ex-plained only on the basis of ambiguities in the corpus of utterances from rvhichthe nerr'rules are induced.2e Before we can determine what constitute ambiguitiesin the sense just defined, r\,e must identify the degree of abstractness of the levelof representation used in morphophonemic rules. This rve cannot do by studyingsynchronic descriptions of languages, for rve have no way of choosing amongequally adequate descriptions employing phonological representations of differentdegrees of abstraction, except in terms of criteria of evaluation which we assuMEto bc correct. What lve must study is the process of induction. Our most abundantand-as our remarks in $6.2 suggest-most reliable source of information forthi-q is inductive change.

Tiie method to be used in studying inductive change is the one we have usedthroughout this investigation, but for rvhich rve have not been able to give theo-retical justification until now. Since inductive change is covert, it cannot beobserved directly. But rve can infer an earlier inductive change from everyattested instance of deductive change. If this method is applied consistently, itwill enable us not only to explain individual instances of morphophonemic change,but al-qo to exploit morphophonemic change zrs evidence for grammatical struc-ture.

REFERENCtrS

Axunnsux, HnxNrNG. 1969a. 'Ihe change of *ol to *od-a Central Slavic innovation. Toappear in Welt der Slaven 14.1969b. Indo-European voicing sandhi in lJkrainian.'Io appear in Scando-Slavica 15.1969c. Lenition in Common Slavic. L9.45.553_74.

2s l{utatis mutandis, a completely parallel statement can be made regarding inductive

and deductive change in phonology. The'mutanda' define the essential differences between

phonology and morphophonemics (cf. Andersen 1969d).zs 'Ambiguities' here means 'distribLrtional regularities which can be produced by alter-

nat ive (sets of) ru les ' .

830 LANGUAGE, voLUME 4b, NUMBER 4 (1969)

1969d. Neutralization and vai'iation. 'I'o appear.ANnnusvsunx, C. H. 1957. Ukrainian-English Dictionary. Toronto: Universit-t, of

Toronto Press.Bnvznnro, S.P. 1960. Istorydna morfolohija ukrajins'koji movy. UZhorod: Oblvydav.Brzpel'ro, O.P.; N'I. K. Bo.rdur; M.A.Zov'ronnlux; S. P. Seltr,rr,nNKo; ri,nd [. J.

TaneNnxro. 1957. Istorydna hramatyka ukrajins'koji movy. Kiev: Radjarrs'ka5kola.

Bur,exovs'KyJ, L. A. 1940. Issledovanija v oblasti grammatideskoj analogii i rodstvennyxIjavlenij. Ifdenye Zapiski Xar'kovskogo Urriversiteta 19 : 1-31.

1951. I(urs sudasnoji ukrajins'koji literaturnoji movy. Kiev: Radjans'ka Skola. I

1956a. Grammatideskaja indukcija v slavjanskom sklonenii. Voprosy Jazyko-znanija 1956: 4.I4-BI.

1956b. Pytannja poxodZennja ukrajins'koji movy. Kiev: Vyd-vo AN URSR. ,

Cnousrv, Noeu, and i\{onnrs H.q.Lr,Fr. 1968.'l'he sound pattern of trnglish. Neiv York:Harper & Row.

Drcls, Paur,. 1932. Aitkirchenslavische Grammatik. Heidelberg: Carl Winter.Her,ln, Monnrs. 1959. The sound pattem of Russian. The Hague: Nlouton.

1963. On cyclically ordered rules in Russian. American Contributions to the FifthInternational Congress of Slavists, 113-32. The Hague: N'Iouton.

Hnrwdnxro, B. D. 1909. Slovar' ukrainskogo jazyka. Kiev.Hnuxs'rvJ, \{. K., and P. K. Kovar,'ov. 1941. Narysy z istoriji ukrajins'koji rnovy.

L'viv : Ukrajins'ke Vyd-vo.JeroesoN, Rouen. 1948. Russian conjugation. Word 4.155-67.

L962. Selected writings, I: phonological studies. The Hague: Nfouton.Kunvrowrcz, Jnnzv. 1960. La nature des procbs dits 'analogiques'. Esquisses linguis-

tiques, 66-86. lVroclarv-Krak6w: Wydawnictwo PAN.1968. The notion of morpho(pho)neme. Directions for historical linguistics, ecl. by

\V. P. Lehmann and Yakov nfalkiel, 65-81. Austin: University of Texas Press.Kvnvdrxro, I. N. 1953-63. Ukrajins'ko-rosijs'k1'j slovnyk, I-VL Kiev: \r.l'd-vo AN

URSR.LnHn-Sprawrr(srr, T. 1965. Wy'b6r tekst6rv do historii jgzyka rosyjskiego. Warsaw:

PWN.LrcurNnn, T. 1\{. 1965. O ciklideskix pravilax v russkom sprjaZenii. Voprosy .Iazyko-

znanija 1965:2.45-54.Mefrczer, Wrror,o. 1958. Tendances gdn6rales des changetnents analogiqtres. Lingua

7 .298-325, 387-420.Mnovnunv, F. P. 1964. Narysy z ukrajins'koji istorydnoji hramatyky. Kharkiv: Vyd-vo

Xarkivs'koho DerZavnoho Universytetu.Mnrr,lnr, AnrorNn. 1964. Introduction n l'6tude comparative des langues indo-

europ6ennes. University, Alabama: IJniversity of Alabama Press.Scuurnr, Hener,n. 1955. Die Rolle der Analogie in der Sprache dargestellt arn Franzdsi-

schen. (Inaugural-Dissertation.) Erlangen.SnuzNovsxr.r, I. I. 1958. Materialy dlja slovarja drevnerusskogo jazyka,I-Ill. Nloscow.

1968. Index a tergo do Material6w do slorvnika staroruskiego. Compiled by I.Dulewicz and others. Warsaw: PWN.

Vasunn, Mex. 1953-57. Russisches etymologisches Worterbuch. Heidelberg: OarlWinter.

Zro*o***rl, V. l,I. 1954. Vnutrennie zakony razvitija jazyka i problerna gramnrati-deskoj analogii. Trudy Instituta JazTrkoznanija AN SSSR 4.74-92.

Zoutor*oX, M.A. 1958. Sekundarnyi i v pryjmennykax i prefiksax ukrajins'koji movy.Sbirnyk naukovyx prac', f, Movoznavstvo, 30-4. Kiev: Vyd' M-vo Osvity URSR.

L962. Pro dejaki javy5da hramatydnoji analohiji v ukrajins'kij movi. Pytannjaistorydnoho rozvytku ukrajins'koji movy, ed. by I. K. Bilodid, 193-209. Kharkiv:Vyd-vo Xarkivs'koho DerZavnoho flniversytetu.

[Received 11 April 1969]