UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Japanese lexical phonology and morphology Ross, Martin John Elroy 1985-06-09

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
831-UBC_1985_A8 R67.pdf [ 3.16MB ]
Metadata
JSON: 831-1.0096517.json
JSON-LD: 831-1.0096517-ld.json
RDF/XML (Pretty): 831-1.0096517-rdf.xml
RDF/JSON: 831-1.0096517-rdf.json
Turtle: 831-1.0096517-turtle.txt
N-Triples: 831-1.0096517-rdf-ntriples.txt
Original Record: 831-1.0096517-source.json
Full Text
831-1.0096517-fulltext.txt
Citation
831-1.0096517.ris

Full Text

JAPANESE LEXICAL PHONOLOGY AND MORPHOLOGY By MARTIN JOHN ELROY ROSS B.Sc, The University of British Columbia, 1977 M.Ed., The University of British Columbia, 1980 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS in THE FACULTY OF GRADUATE STUDIES Department of Linguistics We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA © September, 1985 copyright: Martin John Elroy Ross, 1985 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department Of Linguistics The University of British Columbia 1956 Main Mall Van couve r, Canada V6T 1Y3 Date /HI Abstract Over the years, phonologists working in the generative framework have encountered a number of persistent problems in their descriptions of Japanese phonology. Several of these problems concern phonological rules that sometimes do and sometimes do not apply in seemingly identi cal phonological environments. Many of the proposed analyses achieve observational adequacy, but, nonetheless, are intuitively dissatisfy ing. The first of two such problems involves the desiderative suffix -ta and the homophonous perfective inflection -ta, both of which attach to verb roots. When the verb root is vowel-final, the derivations are straightforward. (1) (a) tabe + ta + i > tabe-ta-i 'want to eat' mi + ta + i > mi-ta-i 'want to see' (b) tabe + ta > tabe-ta 'ate' mi + ta > mi-ta 'see (past)' Derivations are not so straightforward when the verb root is consonant-final. In such cases an intervening i_ is inserted between the root and the desiderative suffix, but not between the root and the perfective inflection. (2) (a) tat + ta + i > tat-i-ta-i 'want to stand' kat + ta + i > kat-i-ta-i 'want to win' (b) tat + ta > tat-ta 'stood' kat + ta > kat-ta 'won' McCawley (1968) is not specific in how he accounts for this differen tial insertion of _i in these phonological identical environments, but ii it appears that he favours the adoption of a morphological rule such as (3) (from Koo, 1974). (3) 0 > i / C]v +tai Koo (1974) has attempted to reanalyze the desiderative suffix as -ita, but, since there is no evidence of W cluster simplification in the language, he is left with the even more difficult problem of delet ing the initial i_ of the suffix following vowel-final verb roots. (4) tabe + ita + i > tabe-ta-i 'want to eat' mi + ita + i > mi-ta-i 'want to see' Maeda (1979) has chosen a boundary solution, positing that t-initial inflections are joined to verb roots by morpheme boundaries (+), while other suffixes such as the desiderative suffix are joined by a stronger boundary (:). By making the i_ insertion rule sensitive to boundaries of level :, the correct outputs can be derived. This solution, though, is unsatisfactory since the assignment of boundaries is not independent ly motivated. A second difficulty encountered by McCawley (1968) and others involves a high vowel syncopation rule that deletes the final i or u of Sino-Japanese monomorphemes when the initial consonant of a following Sino-Japanese monomorpheme is voiceless. (5) iti + too > it-too 'first class' roku + ka > rok-ka 'sixth lesson' However, a morpheme- or word-final high vowel at the boundary between a Sino-Japanese compound and a Sino-Japanese monomorpheme does not delete under those conditions. iii (6) zi-ryoku 'magnetism' (X-Y) zi-ryoku + kei > zi-ryoku-kei 'magnetometer' (X-Y-Z) hai-tatu 'delivery' (Y-Z) betu + hai-tatu > betu-hai-tatu 'special delivery' (X-Y-Z) McCawley accounts for this pattern by invoking internal boundaries of different strengths: + and _#. (7) iti + too roku + ka zi + ryoku # kei betu # hai + tatu He claims, then, that high vowel syncopation is sensitive to boundaries of strength + and is, therefore, blocked from applying to the u of zi + ryoku # kei. His analysis is correct, but his assignment of boun dary strengths is rather arbitrary. Analyses such as the two above which appeal to boundary strength hierarchies have often been intuitively dissatisfying because of a lack of independent motivation. The relatively recent theory of lexical morphology and phonology as formulated by Kiparsky (1982) is ideally suited for this type of problem. One of the theory's most compelling attributes is that phonological processes may be put into a much broad er context that includes morphological processes as well. This more in tegrated approach is often able to fit formerly isolated facts into a network of related facts to provide compelling independent motivation for diverse processes. The purpose of this thesis, then, is to fit i insertion, high vowel syncopation, and other Japanese phonological pro cesses into the lexical phonology network to see exactly how they are related to each other and to the morphological phenomena of the lan guage. iv .3.0..£.^ll«d}&:.'.$8S' Patricia A. Shaw Thesis Supervisor v ( Table of Contents Abstract Table of Contents Acknowl edgement Chapter I: Introduction 1.1: Theoretical Outline 1.2: Source of Data Chapter II: Lexical Morphology and Phonology Notes for Chapter II Chapter III: Phonological Representation of the Syllable Chapter IV: Japanese Syllable Structure Chapter V: High Vowel Syncopation: Evidence for Levels I Chapter VT: Accent System: Further Evidence for Levels I and II Notes for Chapter VT Chapter VII: - 'si, - 'ka, and -teki: Further evidence for Levels I and II Notes for Chapter VII Chapter VTII: Stem Formatives and s/r Deletion VIII.1: Stem Formative i VIII.2: s/r Deletion VIII.3: Level II Compounding Revisited Notes for Chapter VIII Chapter IX: Level III . .11 . .vi ..viii . .1 ..1 ..4 ..5 . .9 . .10 ..15 and II 20 V.l: High Vowel Syncopation 20 V.2: Generative Analysis of High Vowel Syncopation 24 V.3: Lexical Phonology Analysis of High Vowel Syncopation 28 Notes for Chapter V 31 .. .32 ...38 . .39 . .46 . .47 ..47 . .52 . .55 . .57 . .58 vi IX.1: Stem Formative Truncation . 59 IX.2: Infinitives/Connectives ....60 IX.3: Real Epenthesis ....62 IX. 4: s/r Deletion 63 IX.5: Accent ....64 Notes for Chapter IX 67 Chapter X: Conclusion ....68 Bibliography ....70 vii AdaK>wledgenen.t My special thanks are extended to.Dr. Patricia A. Shaw, my thesis supervisor, whose insightful comments based on her impressive knowledge of the field made this thesis possible. Also, I greatly appreciate her patience and encouragement that saw me through the inevitable low spots along the way. Despite the help I received from a variety of people, responsi bility for any errors, omissions, or other weaknesses in this work rests solely on my own shoulders. viii Chapter I Jjitroduction 1.1: Theoretical Outline Over the years, phonologists working in the generative framework have encountered a number of persistent problems in their descriptions of Japanese phonology. Several of these problems concern phonological rules that sometimes do and sometimes do not apply in seemingly identi cal phonological environments. Many of the proposed analyses achieve observational adequacy, but, nonetheless, are intuitively dissatisfy ing. The first of two such problems involves the desiderative suffix -ta and the homophonous perfective inflection -ta, both of which attach to verb roots. When the verb root is vowel-final, the derivations are straightforward. (1) (a) tabe + ta + i > tabe-ta-i 'want to eat' mi + ta + i > mi-ta-i 'want to see' (b) tabe + ta > tabe-ta 'ate' mi + ta > mi-ta 'see (past) ' In (la), the final i. is the indicative, adjectival inflection (see Chapter IX). Derivations are not so straightforward when the verb root is consonant-final. In such cases an intervening i is inserted between the root and the desiderative suffix, but not between the root and the perfective inflection. 1 (2) (a) tat + ta + i —;> tat-i-ta-i 'want to stand' kat + ta + i > kat-i-ta-i 'want to win' (b) tat + ta > tat-ta 'stood' kat + ta > kat-ta 'won' McCawley (1968) is not specific in how he accounts for this differen tial insertion of i in these phonological identical environments, but it appears that he favours the adoption of a morphological rule such as (3) (from Koo, 1974). (3) 0 > i / C]y +tai Koo (1974) has attempted to reanalyze the desiderative suffix as -ita, but, since there is no evidence of W cluster simplification in the language, he is left with the even more difficult problem of delet ing the initial i of the suffix following vowel-final verb roots. (4) tabe + ita + i > tabe-ta-i 'want to eat' mi + ita + i > mi-ta-i 'want to see' Maeda (1979) has chosen a boundary solution, positing that t-initial inflections are joined to verb roots by morpheme boundaries (+), while other suffixes such as the desiderative suffix are joined by a stronger boundary (:). By making the i. insertion rule sensitive to boundaries of level :, the correct outputs can be derived. This solution, though, is unsatisfactory since its main motivation is merely that t-initial inflections behave differently from t-initial derivational suffixes with respect to i_ insertion. Hence, the proposed weaker boundary before t-initial inflections serves as little more than a diacritic exempting such suffixes from i. insertion. 2 A second difficulty encountered by McCawley (1968) and others involves a high vowel syncopation rule that deletes the final i or u of Sino-Japanese monomorphemes when the initial consonant of a following Sino-Japanese monomorpheme is voiceless. (5) iti + too > it-too 'first class' roku + ka > rok-ka 'sixth lesson' However, a morpheme- or word-final high vowel at the boundary between a Sino-Japanese compound and a Sino-Japanese monomorpheme does not delete under those conditions. (6) zi-ryoku 'magnetism' (X-Y) zi-ryoku + kei > zi-ryoku-kei 'magnetometer' (X-Y-Z) hai-tatu 'delivery' (Y-Z) betu + hai-tatu > betu-hai-tatu 'special delivery' (X-Y-Z) Like Maeda (1979), McCawley accounts for this pattern by invoking in ternal boundaries of different strengths: + and _#. (7) iti + too roku + ka zi + ryoku # kei betu # hai + tatu He claims, then, that high vowel syncopation is sensitive to boundaries of strength + and is, therefore, blocked from applying to the u of zi + ryoku # kei. His analysis is correct, but as will be shown later, his assignment of boundary strengths is rather arbitrary. Analyses such as the two above which appeal to boundary strength hierarchies have often been intuitively dissatisfying because of a lack of independent motivation. The relatively recent theory of lexical morphology and phonology as formulated by Kiparsky (1982) is ideally 3 suited for this type of problem. One of the theory's most compelling attributes is that phonological processes may be put into a much broad er context that includes morphological processes as well. This more integrated approach to the grammar of a language is often able to fit formerly isolated facts into a network of related facts to create a convincing whole that can provide compelling independent motivation for any number of different processes. The purpose of this paper, then, is to fit i insertion, high vowel syncopation, and other Japanese phono logical processes into the lexical phonology network to see exactly how they are related to each other and to the morphological phenomena of the language. 1.2: Source of Data Data used throughout this paper were drawn from a variety of sources. By far the most useful sources were: (1) Takenobu, Y. 1940. Kenkyusha's New Japanese-English Dictionary, and (2) Martin, S.E. 1975. A Reference Grammar of Japanese. Lesser amounts of data were obtained from Block (1946), Kageyama (1982), McCawley (1968, 1977), and Parker (1939). In addition to these published sources, several examples were drawn from field notes based on utterances supplied by Mrs. Keiko Shibanuma, a native speaker of the Tokyo (standard) dialect. 4 Chapter II Lexical Morphology and Phonology The basic insight of lexical morphology and phonology is that the lexicon is divided into a series of levels, each with its own component of morphological and phonological rules. The generalized diagram in (8) shows the routing that a basic, underived lexical item may take on its way to Decerning a fully derived and phonologically well-formed word. ;.8) (from Kiparsky, 1982) lexicon underived lexical items level I morphology < > level I phonology / / / level II morphology < > | level II phonology / / / / / / level n morphology | < > [ level n phonology / / / L/ syntax < > postlexical phonology According to (8), basic lexical items are fed directly into the phonological component of level I. Rules of syllabification and metri-5 cal structure (structure-building rules) may apply immediately at this point, if such rules are in fact found at level I in the particular language under discussion. Structure-changing rules of level I are blocked by the Strict Cycle Condition from applying to these forms that are, as yet, nonderived. From level I phonology, the items proceed to level I morphology. If and when a morphological rule is applied to an item, that item is fed immediately back to level I phonology where the newly derived form is run through all the phonological rules of that level to see if any apply. At the end of the battery of phonological rules, that cycle of level I is said to be complete. Any internal brac kets remaining at this stage are erased.-*" The item is then returned to level I morphology where the next cycle begins. Lexical items are cyc led back and forth in this manner between the phonological and morpholo gical components of level I until no more rules apply and the items exit level I. Kiparsky (1982) claims that the outputs of each level must con stitute well-formed lexical entries. In the case of Japanese nouns, it is true that only freely occurring nouns ever emerge from a level. How ever, Kiparsky's claim is untenable for Japanese verbs and adjectives, which are not well-formed until at least one inflection is attached to the stem. These indispensable inflections are not available until late in the grammar. This means that verbal and adjectival outputs from low er levels consist of a stem only and are not able to stand on their own as words. It will be assumed throughout this paper, then, that the ver bal and adjectival outputs of each level need not be well-formed lexical entries, at least as far as the presence or absence of inflections is concerned. 6 There is some discussion in the literature about how the levels are linked — whether the link goes from the phonology of one level to the phonology of the next level, or whether it goes from phonology to morphology (as shown in (8)). Kiparsky (1982) argues in favour of the former possiblity. However, inconsistencies in his argumentation, a full discussion of which is beyond the scope of the present work, sug gest that the latter alternative is the correct one. Indeed, subsequent work of Kiparsky's (1985) reverts to the earlier direction of flow mod elled in (8), albeit without explicit discussion of the issue. Follow ing Kiparsky (1985), therefore, I will assume in this paper a phonology to morphology link. From level I, then, items advance directly to the morphological component of level II and are cycled back and. forth through that level and, subsequently, through all the later levels in the lexicon. Upon emerging from the lexicon, the items are fed into the syntactic compo nent of the grammar and, from there, into the post-lexical phonological component. In contrast to lexical rules, post-lexical rules are noncyc-lic and are sometimes called post-cyclic rules. They are characterized by exceptionless, across-the-board application. Also, they are exempt from the Strict Cycle Condition and, unlike cyclic rules, may apply in a structure-changing function in nonderived environments. There are no published accounts of lexical morphology and phono logy treatments of Japanese grammar. However, a level-ordered descrip tion of Japanese word formation has been worked out by Kageyama (1982) and is a useful reference. Recently, an unpublished Ph.D. thesis has been completed on the subject by Grignon (1985).^ The theory has been 7 successfully applied with interesting results to Malayalam (Mohanan, 1981), Spanish (Harris, 1982), English (Rubach, 1984), and Dakota (Shaw, 1985). 8 Notes for Chapter II: 1. Kiparsky (1982) argues for a weaker Bracket Erasure Convention which applies only at the end of each level. There is no crucial evidence in this thesis for the weaker convention, so the stronger version, which applies at the end of each cycle, is adopted here. 2. Unfortunately, the Grignon thesis was not available at the time this thesis was being prepared. 9 Chapter III Phonological Representation of the Syllable The phonological component of Japanese grammar includes many syl lable sensitive rules that are best characterized by a three dimensional representation of the syllable. Background information on that repre sentation is provided in this section. Three dimensional phonology represents an area of overlap between autosegmental phonology and metrical phonology. The contribution of autosegmental phonology is that, within that framework, the lexical rep resentation of morphemes consists of a segmental tier with its individu al members mapped onto an independent skeletal tier. One conception of the skeletal tier sees it composed of strings of C and V positions (9), a conception that works well in analyses of Arabic (McCarthy, 1981) and reduplication (Marantz, 1982; Yip, 1982). (9) Skeletal Tier C V C C I I I I Segmental Tier bets baits The purpose of the skeleton in such analyses is to act as a relatively stable "backbone" to ensure that correct temporal structure is main tained independently from whatever processes take place in the segmental tier. An important objection to the autosegmental approach is that, despite the supposed independence of the two tiers, there remains an element of redundancy between the two. In particular, since it is ex plicitly required that [+syllabic] segments be matched with V positions ([+syllabic]) on the skeleton and [-syllabic] with C positions ([-sylla-10 bic]) (Marantz, 1982), the feature [+syllabic] can be predicted, one tier from the other. Kaye and Lowenstamm (1982) argue that this redun dancy can be eliminated by reducing the CV positions to bare skeletal points (x) unspecified for the feature [+syllabic]. (10) Skeletal Tier XXX Segmental Tier bet 'bait' They further argue that [+syllabic] information which is unquestionably important in most analyses is recoverable from a third tier, the proso-dic tier, which is a contribution from metrical phonology. (11) $ / \ Prosodic Tier / \ / / \ Skeletal Tier XXX Segmental Tier bet 'bait' To best understand the workings of the prosodic tier, it is necessary to start with a lexical entry and trace in detail the estab lishment of syllable structure, the most relevant aspect of prosodic structure as far as this thesis is concerned. Starting with a lexical entry such as (10), rules of syllabification attempt to syllabify the string. A useful analogy in this regard is to imagine that the skeletal positions are portholes in the side of a ship. Rules of syllabification use these portholes to peer into the ship to see what segments lie with in. Depending on what segments are discovered, a unique syllabification based on universal principles is erected. In this case, b, e, and t are discovered and a syllable is erected with the structure shown in (12). 11 (12) $ / \ / \ / rime / A / / \ / / \ onset nucleus coda X b X I e X t Throughout this thesis, (12) will be simplified to (13) (13) $ / \ / \ 0 = onset / / \ 0 N C 1 1 1 N = nucleus 1 1 1 X X X C = coda bet If the segments b, a, and y_ had been discovered, the syllable structure erected would have been that shown in (14a) or, simplified, (14b). (141 (a) $ / \ / \ / rime (b) / / onset X b nucleus / \ / \ X X $ / \ / \ 0 N 1 / \ XXX I I I bay As a consequence of syllabification, the onset, nucleus, or coda status of each skeletal position is established. Since only consonants can occupy onset or coda positions and only vowels can occupy nucleus 12 positions, it is indirectly determined which skeletal positions are "C" and which are "V". In this manner, then, [+syllabic] information about skeletal positions is recoverable, even though the positions themselves are unspecified for that feature. In a three dimensional (or three tiered) representation, the terminal nodes of the syllabic hierarchy correspond to positions on the skeletal tier rather than to segments, as has more traditionally been supposed. The advantage of the three dimensional version is that seg ments are freed from their one to one correspondence with terminal nodes. This new independence permits configurations such as those in (15) to occur, which characterize geminate consonants (15a), long vowels (15b), and complex segments (15c). (15) (a) X X (b) X X (c) X \ / \ /  \ C V c c Notice that the phonetic length of a segment is determined by the number of skeletal positions or points associated with it. The independence between the segmental and skeletal tiers permits unassociated or floating segments to exist (16). (16) $ $ I / \ N 0 N XXX 'causative suffix' (Japanese) Floating segments are not associated with "portholes" and hence are invisible, at least initially, to rules of syllabification. However, if 13 "wrinkles" in the syllable contour persist after initial syllabifica tion, the floating segment may be provided with a skeletal point and may participate in syllabification if its presence is required. Several examples of floating segments will be encountered in later chapters. The segmental and skeletal tiers are also independent in the sense that changes introduced by rules to one tier do not necessarily effect changes in the other. By the same token, though, rules can exist that do introduce change simultaneously to both tiers. 14 Chapter IV Japanese Syllable Structure The unmarked syllable shape of Japanese is the universally un-A marked C V (skeletal tier not shown). Other possible syllable shapes are: (17) (a) $ (b) $ (c) $ $ I / \ / \ / N  \  \ / I ON / / \ / X I / \ 0 N C 0 I XXX III! V III X X X X (C) V V I I I I (C) V C C Of particular relevance to this thesis is the nature of the coda in Japanese syllables (see 17c).. ..First of all, there are severe res trictions governing which consonants can occupy the coda position. Dealing first with word-final position, the only consonant that can apparently stand as a coda is n. (18) preliminary version $ / \ / \ / / \ 0 N C 1 I I XXX b u n]N 'part' Even this n has a tenuous existence as a coda. Whenever alternate syllabification is possible, the n is whisked out of the apparent coda position and into an onset position. 15 (19) bun + ei > bun-ei 'separate camp' $ / \ $ / \ / \ / / \ 0 N 0 N N 1 X X X X 1 X b u e Normally a full-fledged coda persists as a coda throughout subsequent resyllabifications. In other words, previously established syllable structure tends to be respected. The half-hearted nature of the word-final coda n suggests that it may not be a coda at all. It is proposed here and substantiated further in Chapter VIII that word-final conso nants are actually extrametrical and are largely invisible to rules of syllabification. The correct version of (18), then, is found in (21). (20) extrametricality X | > [ex] / _ C (21) $ / \ 0 N X X I I u n]N part [ex] Despite the syllable structure established in (21), if bun is used independently as a word, the n must be pronounced and, hence, must be incorporated in some way into the syllable structure of the language. Following earlier proposals, then, it is theorized that the potential coda n becomes loosely attached to the syllable as an appendix (A) and X b 16 retains its extrametricality. (22) $ / \ / \ / / 0 N A XXX In contrast to word-final position, genuine codas are found in medial position wherever an onset directly follows. (23) bun + san > bun-san 'break-up' $ / \ / \ $ / / \ / \ 0 N C 0 N X X X X X b u n s a [ex] Throughout the language it can be seen that there is a close relation ship between coda and following onset. In fact, almost the entire set of features of the coda is copied directly from the onset. The only independent feature possessed by the coda is nasality. This means that a nasal consonant in coda position will retain its nasality regardless of the nasality of the following onset. However, all the features of place will be copied from the onset, a process which creates a homorga-nic nasal coda. (24) bun + betu > bum-betu 'separation' bun + kai > bun-kai 'dissociation' 17 Where the coda is non-nasal, its set of features is entirely copied from the onset (total assimilation) and a geminate cluster is produced. (25) (a) kaw + ta —-> kat-ta 'buy (past)' (b) wakar + tari > wakat-tari 'understand (alternative)' The dependency of the coda on the following onset suggests that the onset in some sense governs (Kaye, Lowenstamm, and Vergnaud (1985)) the coda. Furthermore, since no genuine codas exist in the language without governing onsets, it is reasoned that a coda government principle exists. (26) Coda Government Principle: Only those consonants that are governed directly by following onsets may be syllabified as codas. With the exception of nasality, the features of the coda consonant are copied directly from the gover ning onset. To illustrate the action of the Coda Government Principle, a complete derivation of (25a) is presented. (27) kaw + ta > kat-ta (from (25a)) (preliminary version) (a) extrametricality, syllabification $ / \ / \ / / ". 0 N A XXX 1 I I k a w]v [ex] The appendix (A) cannot be analyzed as a coda since there is no follow ing onset present to govern the coda position. 18 (27) (b) inflection, loss of extrametricality, resyllabification $ / \ • / \ $ / / \ / \ 0 N C 0 N X X X X X 1 I I i I k a w t a]y In (27b) the appendix of (27a) is reanalyzed as a coda following the introduction of the onset (t) which can govern the coda position. By the coda government principle (26) the features of the coda consonant are copied directly from the governing onset (t) to produce a geminate cluster (tt). (27) (c) coda government principle $ / \ / \ $ / / \ / \ 0 N C 0 N 1 I I I I X X X X X I I I I I k a t t a]v By the obligatory contour principle, (27c) automatically reduces to (27d). (27) (d) obligatory contour principle $ / \ / \ $ / / \ / \ 0 N C 0 N X X X X X 1 I \/ I k a t a]y 19 Chapter V High Vowel Syncopation: Evidence for Levels I and II V.l: High Vowel Syncopation Over the centuries, Japanese has borrowed a large number of mor phemes from Chinese. The shape of these Sino-Japanese morphemes is gov erned by strict morpheme structure constraints. Specifically, the mor phemes must be from two to four segments in length and must conform to one of the patterns described in (28). (28) {V } (C)V({N }) {CV} Of particular interest are the morphemes of the shape (C)VCV, composed of two light syllables. In these three and four segment morphemes the final pair of segments must be a member of the following set: ki, ku, ti, or tu^. A widespread rule of high vowel syncopation deletes the final high vowel of such pairs when the next morpheme is also Sino-Japanese and begins with a voiceless consonant . (29) High Vowel Syncopation X I —> 0 / C ]N C V [+Sino] [+Sino] [+hi] [-vcd ] [-vcd ] As indicated, the application of the rule deletes the high vowel toge ther with its associated skeletal point — an example of a rule that acts simultaneously on two tiers. 20 (30) teki + koku > tek-koku 'enemy country' (a) Sino-Japanese compounding, syllabification $ $ $ $ / \ / \ / \ / \ 0 N 0 N 0 N 0 N X 1 X X X X X X 1 X 1 t e k k o 1 k (b) high vowel syncopation $ $ $ / \ / \ / \ 0 N 1 0 1 N 0 N X 1 X X 1 X I X X 1 X t e 1 k o 1 k (c) resyllabification $ / \ / \ $ $ / / \ / \ / \ 0 1 N C 1 1 0 N 1 1 0 N 1 1 1 X 1 I X X 1 1 X X 1 1 1 1 X X t 1 1 e k]N 1 1 k o k u]N In (30c), the k is permitted to be resyllabified as a coda because it is governed by a following onset. (d) obligatory contour principle $ / \ / \ $ $ / / \ / \ / Y 0 1 N C 0 N 1 1 1 1 0 N 1 1 1 X 1 1 1 1 X X X X 1 1 X X 1 1 \ / 1 1 1 t e k o k u] N 21 (31) more applications of high vowel syncopation roku + ka > rok-ka 'sixth lesson' iti + too > it-too 'first class' butu + tai > but-tai 'solid object' High vowel syncopation does not usually occur when either or both of the elements are native. (32) miti + kusa > mitikusa 'loiter' (road + grass) [+nat] [+nat] ' kutu + sita > kutusita 'socks' (shoe + under)\ [+nat] [+nat] The deletion of the high vowel in the examples of (30) and (31) creates consonant clusters that are permitted in Japanese. That is not always the case. [33) (a) iti + sen •> *itsen 'one thousand' (b) iti + ki > *itki 'first period' (c) butu + situ > *butsitu 'substance' The underlined clusters in (33) are not permitted in Japanese because they are in violation of the coda government principle (26), as revealed in (34). (34) butu + situ > bus-situ (33c) (a) Sino-Japanese compounding, syllabification $ $ $ $ / \ / \ / \ / \ 0 1 N 0 1 N 1 1 X 1 X 1 X 1 X 1 s i t ii] U t UJn o UjN 22 (b) high vowel syncopation $ $ $ / \ / \ / \ 0 1 N 0 N 1 X 1 X X 1 X X 1 1 b u 1 t; lN s 1 i u] N The isolated t in (34b) is resyllabified as a coda and, by the coda government principle, takes on the features of its governing onset and becomes an s. (c) coda government $ / \ / \ $ $ / / \ / \ / \ 0 N C 0 N 0 N X X X X X X X b u s]N sit u]N (d) obligatory contour principle $ / \ / \ $ $ / / \ / \ / \ 0 N C 0 N 0 N 1 I I I I I I X X X X X X X II \/ III b u s it u]N The'correct derivations for the words in (33), then, result from the ap plication of high vowel syncopation, the coda government principle, and the obligatory contour principle. (36) iti + sen > is-sen 'one thousand' iti + ki > ik-ki 'first period' butu + situ > bus-situ 'substance' 23 V.2: Generative Analysis of High Vowel Syncopation The story of high vowel syncopation as described above would be neat and tidy if it were not for other facts. First of all, some high vowels that are in a position to be deleted by (29) are immune to the rule. (39) hai + tatu > hai-tatu 'delivery' (Y-Z) betu + hai + tatu > betu-hai-tatu 'special delivery' (X-Y-Z) sya + zitu > sya-zitu 'realism' (X-Y) sya + zitu + ha > sya-zitu-ha 'realist movement' (X-Y-Z) zi + ryoku > zi-ryoku 'magnetism' (X-Y) zi + ryoku + kei > zi-ryoku-kei 'magnetometer' (X-Y-Z) As noted by McCawley (1968), the exceptionality of these cases is sys tematic. The immediate constituent structure of the three in (39) is as follows: (40) A A A / \ / \ / \ /A A \ A \ / / \ / \ \ / \ \ betu-hai-tatu sya-zitu-ha zi-ryoku-kei [specia1[delivery]] [[realist]movement] [[magnetism]instrument] Clearly, in each example, the vowel that fails to delete is the one that falls at the major constituent break. To account for this pattern, McCawley proposes that high vowel syncopation is sensitive to boundaries of strength + (morpheme boundary) and not to boundaries of strength # (internal word boundary). Thus, the components of (39) betuhaitatu are joined as betu#hai+tatu and the u of betu is not in the appropriate environment to be deleted. McCawley justifies this assignment of boun daries by noting that the constituents of hai-tatu, sya-zitu, and zi-ryoku are bound in the sense that they never stand alone as words. 24 The immediate constituents of the words in (40), on the other hand, are freely occurring words. Hence, the words of (40) are examples of "words embedded within words" which by convention carry internal word boun daries. What McCawley is apparently claiming, then, is that high vowel syncopation can only apply between bound morphemes: Note now that in the [examples in 40] the constituents zi-ryoku, sya-zitu, and hai-tatu are not merely sequen ces of Sino-Japanese morphemes but are indeed words in the sense that these compounds are formed by rules which embed a word within a word (as contrasted with rules which form compounds by joining together two pieces neither of which is a word, as in psychology, telegraph, etc.). According to convention, when a word is aitedded within a word, it carries an internal word boundary with it. (McCawley 1968; 117) McCawley's claim is demonstrably inadequate. There are many instances of high vowel syncopation between one bound and one free morpheme and even between two free morphemes. (41) (a) free + bound (i) tet-tei 'horseshoe' from tetu]p 'steel' and -tei]g 'hoof (ii) tak-ken 'far-sightedness' from taku]p 'desk' and -ken]B 'view' (b) bound + free (i) hit-tan 'stroke of the pen' from hitu]g 'writing' and tan]p originate (ii) bos-syo 'dead letter' from botu]R 'dead' and syo]p 'writing' 25 (c) free + free (i) has-sya 'departure of train' from hatu 'leaving' and sya 'train' (ii) hap-pyaku 'eight hundred' from hati 'eight' and hyaku 'hundred' (iii) gyak-kyoo 'adversity' from gyaku 'contrary' and kyoo 'state' (iv) tak-kyuu 'ping pong' from taku 'table' and kyuu 'ball' The components of the words in (41c) are obvious'examples of "words em bedded within words", and as such should be joined by McCawley's inter nal word boundary (#) and thus be ineligible for high vowel syncopation. That something other than boundary information is involved in the high vowel syncopation story is demonstrated by the behavior of the verb suru 'to do'. Suru is an unusual verb in that it can stand on its own or it can be affixed to verbal nouns to derive verbs. Verbal nouns com prise a special lexical category in Japanese that function in some res pects like nouns and in other respects like verbs (Kageyama, 1982). Many Sino-Japanese elements count as verbal nouns and can participate in suru derivation. Suru is also unusual in that it is a native element that triggers high vowel syncopation. As illustrated in (42) and (43), suru can combine with bound as well as free Sino-Japanese forms. (42) bound + suru (a) mes-suru 'perish' (from metu-]g) (b) tas-suru 'reach' (from tatu-]B) (c) bos-suru 'sink' (from botu-]R) 26 (43) free + suru (a) ris-suru 'measure' (from ritu-]F) (b) has-suru 'discharge' (from hatu-]p) (c) kes-suru 'decide' (from ketu-]p) Suru can also attach to Sino-Japanese compounds. In all such cases suru fails to trigger high vowel syncopation. (44) compound + suru (a) koo-tatu-suru 'notify verbally' (cf. 42b) from koo-tatu 'official announcement' (b) ma-metu-suru 'wear down' (cf. 42a) from ma-metu 'abrasion' (c) sen-ritu-suru 'shudder' (cf. 43a) from sen-ritu 'a shiver' The obvious fact to be accounted for in all these data (40-44) is that, regardless of whether individual elements are free or bound, high vowel syncopation can only apply to the high vowel at the boundary bet ween the innermost two elements. High vowel syncopation will not apply to high vowels adjacent to boundaries created by the addition of third elements. Traditional boundary notation can handle this pattern, but only in a contrived way. (45) (a) metu + suru > mes-suru 'perish' (42a) (b) ma + metu # suru > ma-metu-suru 'wear down' (44b) Since, as noted in (29), the structural description of high vowel synco pation includes a morpheme boundary, the rule operates on (45a) and along with cluster simplification and spreading produces the correct output. The # boundary in (45b) correctly blocks the application of the rule. This analysis, however, is objectionable for one main reason. 27 Evidently, the rank of the boundaries in (45) is determined solely by a rule something like (46) that totally disregards the nature of the ele ments that are being joined together. (46) Assign rank + to the boundary between the innermost pair of elements in a compound. Assign rank # to any other boun daries that may occur. This indiscriminate assignment of rank ignores the fact that hatu and sya (41c), for example, are free forms that would normally be joined as hatu#sya. V.3: Lexical Phonology Analysis of High Vowel Syncopation Consider now how these several facts might be treated in a lexical phonology framework. Since the output from each level must be a lexical entry, bound morphemes cannot survive unoompounded to level II. The fact that Sino-Japanese compounding often involves bound forms has lead Kageyama (1982) to conclude that such compounding occurs at level I in Japanese. One would also want to include Sino-Japanese and suru compounding at level I since suru also attaches to bound Sino-Japanese morphemes (42). If these processes do in fact occur in the morphological component of level I, then high vowel syncopation must occur in the phonological component of level I since the rule is sensitive to the-boundary information bet ween Sino-Japanese elements of compounds. That boundary information is lost through bracket erasure by the time the compound reaches level II. An interesting fact about Japanese is that one never finds com pounds of three bound forms. The third element, if present, is always 28 free. Also, as previously noted, high vowel syncopation never occurs at the boundary created by the third element. It is difficult to imagine how these two facts could be accounted for in any framework other than the one outlined here. Apparently, the compounding of monomorphemes (usually Sino-Japanese morphemes) takes place at level I and all subse quent compounding involving third members must be postponed until level II. As an example of this level-ordered process, the derivational his tories of (42a) and (44b) (see-(45)) are compared in (47). (47) (a) /metu + suru/ (b) /ma + metu + suru/ Level I moncmorphemic compd metu + suru ma + metu high V syncopation 0 coda government s Level II compounding ma-metu + suru mes-suru ma-metu-suru In (47), the u of. ma-metu cannot be deleted at level I since it is not in a suitable environment for high vowel syncopation. At level II, it cannot be deleted, because high vowel syncopation is not available at that level. In the same manner, it is easy to account for the failure of the u of zi-ryoku (39) to delete. 29 (48) Level I monomorphemic compd high V syncopation Level II compounding /zi + ryoku + kei/ zi + ryoku zi-ryoku + kei zi-ryoku-kei 30 Notes for Chapter V: 1. Instances of tu and ku outnumber those of ti and ki by over five to one. 2. The application of high vowel syncopation is more restricted in the case of morphemes ending in ki and ku than it is for those ending in ti and tu. In the former, the high vowel is only deleted when the following consonant is a k. (a) teki + koku > tek-koku 'enemy country' gaku + koo > gak-koo 'school' (b) seki + tan > seki-tan 'coal' toku + hon > toku-hon 'anthology' Of the two, ku morphemes are less problematical that ki morphemes which sometimes display optional high vowel deletion and sometimes do not permit it at all. (c) seki + kan > seki-kan or sek-kan 'sarcophagus' heki + ken > heki-ken 'prejudice' (*hek-ken) teki + ka > teki-ka 'drip' (*tek-ka) 31 Chapter VT Accent System: Further Evidence for Levels I and II The rudimentary level-ordered description developed in the pre ceding chapter helps to explain a very striking pattern involving the accentual system. Syllables of Japanese may carry either a high tone (H) or a low tone (L). Only certain patterns of tones are permitted. For example, amongst trisyllabic nouns that are followed by monosyllabic case mark ers, only four surface patterns are possible. (49) H L L L makura ga 'pillow' L H L L kokoro ga 'heart' L H H L atama ga 'head' L H H H sakana ga 'fish' •Since there is no more than one pitch fall from H to L per word, each pattern can be predicted from the placement of an accent mark (') mark ing that pitch fall. Words without a pitch fall are left unaccented, as in sakana ga (50). (50) ma'kura ga koko'ro ga atama' ga sakana ga All syllables following the accent mark carry L tones, while most of those preceding the accent mark carry H tones. The exception is the first syllable which, in the Tokyo dialect, is L if the following tone 32 is H. Notice that in isolation atama' and unaccented sakana carry iden tical tone patterns. Thus, the accentedness of a word often cannot be determined conclusively from its tone pattern heard in isolation. In contrast to nouns, verbs and adjectives have only two accen tual possiblities regardless of length. To best understand these pat terns it is easiest to think in terms of rnorae rather than syllables. The mora has been lightheartedly defined by McCawley (1977) as "some thing of which a [heavy] syllable consists of two and a [light] syl lable consists of one". Since a heavy syllable consists of what would make up a light syllable, plus additional material, one can take the initial (C)V- of a heavy syllable to be its first mora and the remaining -V or -C to be its second mora. Thus semboo envy' can be divided into two syllables (sem-boo) and four morae (se-m-bo-o). Verb and adjective stems are accented either on the penultimate mora (i.e. the pitch falls after the penultimate mora) or not at all. (52) (a) ta'be-]v 'eat' (51) L H H atama L H H sakana tano'm-] V 'ask for' ake-] V 'open' (unaccentable) susum- ] V 'advance' (unaccentable) (b) ta'ka-] A 'high' kura-] A 'dark' (unaccentable) 33 Concatenations of derivational suffixes may be attached to the verb or adjective stem. The suffixation of each successive morpheme creates a new, longer stem that, like the shorter stems, is accented on the stem penultimate mora. Previously established accents are dropped in favour of the newest, rightmost accent. Concatenations built on un accented stems remain unaccented (53a). In (53) the suffixes in paren theses are present indicative inflections that will be ignored for the time being"'". (53) (a) ake]y+ (ru) > ake-(ru) 'open' ake]v+ sase]v+ (ru) > ake-sase-(ru) 'makes open' (b) tabe]v+ (ru) > ta'be-(ru) 'eat' ta'be]v+ sase]v+ (ru) > tabe-sa'se-(ru) 'makes eat' tabe-sa'se]v+ rare]v+ (ru) > tabe-sase-ra're-(ru) 'makes eat (pas)' (c) yom]v+ (ru) > yo'm-(ru) 'read' yo'm]v+ ta]A+ (i) > yom-i'-ta-(i)2 'want to read' yam-i'-ta]A+ gar]v+ (ru) > yom-i-ta-ga'r-(ru) 'behaves like he wants to read' Relating this section to the previous one, verbs can be derived . from Sino-Japanese monomorphemes, compounds, and other verbal nouns by suru. One might expect that the derived verbs would all be accented on the stem-penultimate mora (i.e. stem]VN+ suru]v > stem-su'ru]^) according to the pattern evident in (53). Yet the expected accent placement only occurs when suru is attached to a monomorpheme (54a). When suru is attached to a compound, the original accent of the compound prevails (54b). 34 (54) (a) a'i]^ ai-su'ru]v 'love' sa'n]^ san-su'ru]v 'produce' (b) se'i-ri]^, se 'i-ri-suru]v 'arrange' *sei-ri-su'ru ki-to']^ ki-to'-suru]v *ki-to-su'ru] V 'scheme' The difference between the (a) and (b) patterns in (54) can easily be accounted for by placing the verbal and adjectival accent rule in the phonological component of level I. (55) level I monomorphemic compd V/A accent level II compounding (a) /a'i + suru/ (b) /se' + iri + suru/ a'i]™ + suru]v ai-su ru]v se'W iri] VN se'-irilyj^i- suru] ai-su ru se -lri-suru -V As shown in (53), the following suffixes provide input to the level I V/A accent rule: -rare (passive), -sase (causative), -ta (desiderative), and -gar (A—>V). Consequently, consider the effects of the concatenation in (56) (from Martin (1975)): 35 (56) tabe]v+sase]y+rare]v+ta]A+gar]y-i-(ru) > tabe-sase-rare-ta-gar-ru 'make him behave like he wants to be eaten' Another suffix, the semblative -rasi 'seems like', can fit into this sequence of suffixes before -gar, e.g. otoko]N~rasi]A~gar]v~(ru) "behaves in a manly fashion', -rasi derives adjectives from nouns (primarily), from a few verbs, and even from a few adjectives. (57) otoko]N~rasi]A-i > otoko-rasi '-i 'man-1ike ' ame]N~rasi]A-i > ame-rasi'-i 'rain-like' ' kitana]A~rasi ]A-i >kitana-rasi'-i 'dirty-looking' niku]A-rasi]A~i > niku-rasi'-i 'hateful-looking' The suffix -kata is a verb nominalizer that can attach to vir tually any verb (derived or nonderived) to derive an abstract noun. The nouns created by -kata are unusual^ in that they seem capable of being accented by the V/A accent rule. Unaccentable verb stems yield unaccen table nouns (58a) and accentable become accentable on the penultimate mora (58b), just like ordinary verbs and adjectives. (58) (a) unaccentable verb stems ake]v+ kata]^ > ake-kata 'way to open' umare]v+ kata]N > umare-kata 'way of being born' (b) accentable verb stems mi]y+sase]v+ kata]N > mi-sase-ka'ta 'way of causing to see' kawai]A+ garJv+ kata]N > kawai-gar-(i)-ka'ta 'way of being loved' Since all the suffixes -sase, -rare, -ta, -rasi, -gar, and -kata . provide input to the V/A accent rule, all of them must be added at the 36 same level as the rule, i.e. level I. If this were not so and -rare, for example, were added at a later level, then *tabe-sa'se-rare-(ru) 'cause to be eaten' would be the expected accentuation. The suffixes of (59) occur at level I. (59) -; sase -rare -ta -rasi -gar -kata (causative) (passive) (desiderative) (semblative) (A—>V) (N,V,A—>N) Up to this point, then, the following morphological and phonolo gical processes have been identified at the first two levels of Japanese grammar. ;60) Level I rnonomorphemic compd derivation: -sase -rare -ta -rasi -gar -kata high vowel syncopation V/A accent Level II compounding 37 Notes for Chapter VI 1. As noted in Chapter II, these inflections are not available until later in the grammar. The minor effects on accent of a few inflec-. tions will be discussed in Chapter IX. 2. The extra i is a stem formative which is automatically suffixed to consonant-final verb stems. However, under certain conditions, the i may not be expressed. See Chapter IX. 3. Suru is an irregular verb, making it impossible to decide exactly what constitutes the verb root and what constitutes the inflection. Since this issue does not affect the analysis, suru is introduced here as a complete unit rather than as two separate morphemes. 4. Normally the accentuation of derived nouns is much less predictable. 38 Chapter VII -'si, -"ka, and -teki: Further Evidence for Levels I and II Using the level ordered description in (60), 'si 'regard as', 'ka '-ize', and teki '-type, -ic, -ical' can now be placed at appropriate levels. Traditionally these morphemes have been analyzed as suffixes (Martin, 1975), but evidence presented in this section challenges that view. The traditional analysis is based mainly on the observation that these morphemes usually occupy positions to the right of lexical mor phemes, slots typically occupied by suffixes. Also, as with suffixes, their function seems primarily grammatical. First consider 'si and "ka. Both attach almost exclusively to foreign nouns and adjectival nouns (mostly Chinese and English origin) to derive verbal nouns. Adjectival nouns (AN), like verbal nouns, com prise a special lexical category in Japanese. They share properties of both nouns and adjectives. Interestingly, all borrowed foreign adjec tives fall into the adjectival noun category (Kageyama, 1982). Martin (1975) suggests that 'si derivatives are bound in the sense that they are usually followed directly by suru. The verbal nouns created by 'ka, on the other hand, may optionally take suru. Both morphemes are lex ically pre-accented which causes the pitch of a derived form to fall after the penultimate mora of the stem. 39 > be's-si 'disregard' from betu- 'contempt' > ge'n-si 'visual hallucination' from gen- 'supernatural' > ko'k-ka 'blackening' from koku- 'black' > ko'k-ka 'ossification' from kotu 'bone' The examples in (61) show that both 'si and 'ka attach to (Sino-Japanese) morphemes (bound and free) and trigger high vowel syncopation. This involvement with high vowel syncopation "firmly establishes 'si and "ka at level I. But there is evidence that these morphemes may also be added at level II. When attached to Sino-Japanese compounds of two or more elements, high vowel syncopation does not occur. (62) (a) toku + betu]AIg+ "si]y^+ suru]y > toku-betu'-si-suru 'regard as special' (from tokubetu 'special') *toku-be 's-si-suru (b) doo + itu]^* "si]yfl+ suru]v > doo-itu'-si-suru 'regard as identical' (from docitu 'sameness') *doo-i's-si-suru (c) gen + zitu]y^+ 'ka]^ > gen-zitu'-ka 'actualization' *gen-zi'k-ka The pattern in (62) is strikingly similar to that involving suru in (44), which indicates that 'si (63) and Tea must be available at level II as well as at level I. (61) betu]B+ 'SOJVN gen]B+ 'si]^ — koku]B+ 'JsaJvN kotu]F+ 'kalyjj 40 (63) Level I S-J compounding Level II 'si attachment compounding Conclusive evidence of the level II nature of 'si is found in (64) where 'si is clearly attached outside a level II compound. (64) (a) kiken-zimbutu 'a dangerous character' (b) kiken-zimbutu'-si-suru 'regard as a dangerous character' (c) *zimbutu'-si-suru (d) kiken'-si-suru 'look askance at' In (64b), 'si must be attached to the compound kiken-zimbutu as a whole, and not to the single word zimbutu, since *zimbutu'-si-suru (64c) is ungrammatical. Since the type of compounding shown in (64a) is a level II phenomenon and since 'si is added after the compounding (i.e. is attached to the compound), 'si must occur at least at level II. It would also appear that lea is attached outside level II com pounds, since 'ka is never found compound-internally (65c). Therefore, 'ka must also be found at least at level II. (65) (a) zyuukagaku-koogyoo 'heavy chemical industry' (b) zyuukagaku-koogyoo-ka "heavy chemical industrialization' (c) *zyuukagaku'-ka-koogyoo There is strong evidence, then, that 'si and "ka are added at both levels I and II. Using the same type of arguments, it can be determined that teki is also a bi-level morpheme, teki attaches to nouns, the vast majority of which are [+foreign] (mostly Sino-Japanese), to derive adjectival nouns. Translated into English, teki imparts a /toku + betu + 'si + suru/ (62a) toku]AN + k^-^AN toku-betu]^ + 'sijyjj toku-betu- 31]^+ suru]y toku-betu- 'si-suru 41 meaning something like "-type, -ic, or -ical". All teki derivatives are unaccented even if the stem originally carried a lexical accent"'". As exemplified in (66a), teki attaches to Sino-Japanese monomorphemes and undergoes high vowel syncopation, establishing it at level I. As expected, high vowel syncopation does not apply after attachment to Sino-Japanese bimorphemes (66b), which indicates that a level II teki is involved. (66) (a) si-teki 'poetic' (from si]N 'poetry') but-teki 'physical' (from butu]N 'material') (b) gi-zutu-teki 'technical' (from gi-zutu]N 'technique') *gi-zut-teki hoo-katu-teki 'inclusive' (from hoo-katu]N 'inclusion' *hoo-kat-teki The fact that teki can be attached to various kinds of elements ranging ing from simple nouns to long compounds such as taisyuu-syoosetu-teki 'popular-novelish' has lead Kageyama (1982) to also conclude that teki is found at level II . So it would seem that 'si, Tea, and teki each occur on at least two different levels. If the -three are in fact suffixes as is commonly supposed, then the situation is theoretically complex. Amongst the lan guages of the world, it is usually observed that each language has a fixed sequence to its suffixes, whereby suffix X always precedes suffix Y (if present) which, in turn, precedes affix Z (if present), and so on. The ordering of the sequence is determined in two ways. Firstly, each suffix is ordered (if necessary) with respect to the other suffixes 42 within its level and, secondly, suffixes of lower levels always precede those of higher levels. Thus, in (67), a permissible suffix sequence is A-C-Z, but not *C-A-Z or *A-Z-C. (67) level I A > B > C level II X > Y > Z Needless to say, if the same suffixes appear at more than one level, then numerous ordering paradoxes are possible. In (68), for example, a sequence such as A-B-A would be possible even though A is ordered before B at level I. (68) level I A > B > C level II A > Y > Z If some suffixes in Japanese have bi-level membership, then one would expect to find ordering paradoxes throughout the language. In actual ity, such paradoxes are at best rare. It is possible that the grammar of Japanese is complex enough to ensure that paradoxes are filtered out before they surface. In this case, though, there is evidence that the three "suffixes" are actually bound lexical morphemes and not suffixes at all. Dealing first with 'si, examination of (61), (62a, b), and (64) reveals certain similarities between the English glosses for 'si "suffixed" words. Most glosses take the form 'regard as X'. For exam ple, in (61a) ge'n-si 'visual hallucination', gen- is more related to to 'hallucination' than to 'visual'. In zu'nsi- 'tour of inspection', zun-means 'tour'. In these two words, then, 'si seems to mean 'visual' and 'inspection' respectively, both of which are related to "seeing". Even 43 regard as X' intuitively involves "seeing", i.e. 'seeing as X'. This, impression is borne out by the words in (69) where 'si is the first member of compounds and, just like the alleged suffix 'si, carries the meaning "seeing, visual, etc.". (69) si-tyoo 'sight and hearing' si'-wa 'lip reading (visible speech)' si'-ya 'field of vision' Incidently, it may also be pointed out that 'si is written in Japanese with the same character no matter whether its position is word-initial or -final. It is hypothesized here, therefore, that the two 'si's (word-initial and word-final) are the same morpheme. Since true suf fixes never occupy a word-initial position, a case can be made that 'si is just an ordinary Sino-Japanese verbal noun that is capable of com pounding with other Sino-Japanese morphemes. As with 'si, 'ka may also appear as the first member of a com pound and can be considered a lexical Sino-Japanese morpheme. An inter esting twist is that, at least in the case of (70), the ordering of the members has no discernable effect on the meaning of the compound. (70) ko'k-ka 'ossification' (from kotu 'bone' + 'ka) ka-kotu 'ossification' (from 'ka + kotu) It is more difficult to prove that teki may occur as the first element of a compound. There are a few instances of compounds that begin with a teki (71) that is written with the same Japanese character as the suffix-like teki of (66), but the problem is demonstrating that the two are one and the same. 44 (71) teki-tyuu 'hit the mark' teki-kaku 'accurate' Neither of the two teki's has a clearly defined semantic content that is carried with it from compound to compound, so a direct comparison of the meanings of the two teki's is problematical. However, since, in the examples I have studied, Japanese characters are reliable "tags" identi fying which morpheme is which, I will assume that the teki of (66) is the same as that in (71). Under that assumption, teki is a lexical morpheme. It can be concluded, then, that 'si, 'ka, and teki are all bound lexical morphemes that may appear at any level that permits compounding (i.e. levels I and II). 45 Notes for Chapter VTI 1. The precise mechanism of this phenomenon is unclear, and a complete investigation is beyond the scope of this thesis. 2. Kageyama apparently rejects the possibility that teki might also be found at level I. However, since his work is based solely on morpho logical data, he has not considered the crucial phonological evidence that has lead me to my bi-level interpretation. 46 Chapter VIII Stem Formatives and s/r Deletion Most verb roots are consonant-final (e.g. kak- 'write') rather than vowel-final (e.g. tabe- 'eat'). Complicated phonological inter actions may result when these root-final consonants are brought adjacent other consonants. Two of these interactions involve stem formatives and s/r deletion. VTII.l: Stem Formative i When t- and k-initial suffixes are attached to consonant-final roots at level I, an intervening i materializes (72b-e). This inter vening i does not appear following a vowel-final root (72a). (72) (a) tabe + ta + (i) > tabe-ta-(i) 'want to eat' (b) tat + ta + (i) >.tat-i-ta-(i) • 'want to stand' (c) kak + ta + (i) > kak-i-ta-(i) 'want to write' (d) kak + kata > kak-i-kata 'way to write' (e) sin + kata > sin-i-kata 'way to die' The nature of this intervening _i has been the subject of considerable debate in the literature. As noted in the introduction to this paper, phonological and morphological analyses have been propounded to account for the insertion, non-insertion, and/or deletion of the i.. Evidence for treating the i_ as epenthetic is weak, since it does not appear to serve a phonological function. If the i_ were serving a phonological function, its presence would reduce the markedness of the system by breaking up unacceptable consonant clusters, smoothing out syllable contours, or maximizing the number of unmarked syllables. Examination of (72), for example, reveals that the function of the intervening i is plainly not 47 to break up unacceptable consonant clusters and thereby maximize the number of unmarked CV syllables. Geminate stops such as the t-t and k-k broken up in (72b) and (72d) respectively, and the n-k cluster broken up in (72e) are all perfectly acceptable. Furthermore, the same inter vening i appears after consonant-final initial members of verb com pounds, regardless of whether the second member of the compound is consonant- (73a) or vowel-initial (73b). —> kak-i-yam-(ru) 'write-stop' -> yob-i-das-(ru) 'call-stop' —> oycg-i-kir-(ru) 'swim-through' -> yart-i-owar-(ru) 'read-finishes' —> kik-i-akir-(ru) "hear-wearies' -> ur-i-isog-(ru) 'sell-busy The data of (73b) show clearly that the purpose of the i_ is not phono logical, since its presence adds to the markedness of the string by creating two nuclei adjacent to each other. It would appear that the only consistent function of the i is a morphological one to act as a stem formative to create vowel-final verb roots in all cases. As will become clearer subsequently, it turns out that a morphological analysis involving a stem formative is best able to handle the observed data. The stem formative i is introduced by rule (74) following each consonant-final verb stem at level I. Once introduced, the stem forma tive has a rather tenuous existence. As will be shown later, if the derivational function of the stem formative is unused, the _i will simply drop. This ephemeral nature suggests the stem formative lacks the per manence of an associated skeletal point. For this reason, the i is 48 (73) (a) kak]v+ yam]v + (ru) -yob]v+ das]v+ (ru) — °y°g]v+ kir]v+ ~ (b) yom]v+ owar]v+(ru) — kik]v+ akir]v+ (ru) -ur]v+ isog]v+ (ru) — introduced as a floating segment (see chapter III] (74) stem formative 0 > i]v / C]v  To see how the stem formative functions, consider the three dimensional analysis of (72e) presented in (75). (75) sin + kata > sin-i-kata (72e) (a) level I, cycle I: extrametricality, syllabification $ / \ 0 1 N 1 X 1 X X 1 s i n] [ex] V (b) level I, cycle II: stem formative $ / \ 0 N XXX s i n]vi]v [ex] It is assumed here that the mere segmental presence of the floating i of the stem formative does not disrupt the extrametricality of the n. That is, since it is the skeletal X points rather than the segments them selves which are prosodically relevant, it may be hypothesized"1- that the skeletal point dominating n will retain its extraprosodic status as long as it is the rightmost skeletal point in the string. Presumably only a segment with an associated skeletal point would precipitate the loss of extrametricality. 49 Rules of syllabification are obligated to syllabify all skeletal points that are visible to them. Once that is accomplished, and if the resultant syllabification is unmarked, then syllabification is complete and the rules are "content". In (75b), an unmarked syllable has been created and the extrametrical consonant and the floating i_ are prosodi-cally invisible. Therefore, syllabification is complete for that cycle. [75) (c) level I, cycle III: bracket erasure, derivation, loss of extrametricality $ $ $ / \ / \ / \ 0 1 N 0 N 1 X X I X X X s 1 i n i]v k a a] N The addition of the new material -kata, forces the loss of extrametrical ity. The formerly extrametrical n is now visible to the rules of syl labification and must be syllabified. In this case, the n cannot be syllabified as a coda, because it cannot be governed directly by the following onset k due to the presence of the intervening _i. With no other options, the rules of syllabification are forced to punch some holes in the side of the ship and look for segments without portholes that might be of assistance. The floating i is discovered, provided with a porthole (point), and is pressed into syllabic service (75d). (75) (d) level I, cycle III: resyllabification $ $ $ $ / \ / \ / \ / \ 0 1 N 1 0 | 1 X 1 X 1 1 X s 1 i n ]v k a t a]N 50 Consideration of a second type of example shows that the level I floating stem formative accounts well for patterns evident at level II where the i appears between elements of verbal compounds (73). The derivational history of (73a) is traced in (76). (76) kak + yam + (ru) > kak-i-yam-(ru) (73a) (a) level I, cycle II: extrametricality, syllabification, stem formative $ / \ 0 N 1 I XXX k a k]y i]y At level I, yam- follows a similar derivational history to kak-. At level II, however, there are some differences which, though illustrated below, will be discussed in detail in chapter IX. (75) (b) level II, cycle I: compounding, loss of extrametricality $ $ / \ / \ ON ON II II XXX XXX' [kak ilyty a m [ex] (c) level II, cycle I: syllabification (cf. (75d)) $ $ $ / \ / \ . / \ 0 1 N 1 0 N 0 N 1 X 1 X 1 X 1 X 1 X X l [I 1 a 1 k 1 . 1 MY 1 a 51 . The examples in (73b) are especially interesting since the second members of the compounds are vowel-initial. They will be examined fur ther in section VIII.3. VIII.2: s/r Deletion When one of the s_- or r-initial suffixes of level I attaches to a consonant-final verb root, the s or r deletes. (77) (a) ake + sase + (ru) > ake-sase-(ru) 'open (caus) ' (b) ake + rare + (ru) > ake-rare-(ru) 'open (pas)' (c) hanas + sase + (ru) > hanas-ase-(ru) 'speak (caus)' (d) isog + sase + (ru) > isog-ase-(ru) 'hurry (caus)' (e) isog + rasi + (i) > isog-asi-(i) 'is busy' (f) itam + rasi + (i) > itam-asi-(i) 'is sad' (g) yorokob + rasi + (i) > yorokob-asi-(i) 'is joyful' (h) wakar + rare + (ru) > wakar-are-(ru) 'understand.(pas) ' (i) sin + rare + (ru) > sin-are-(ru) 'die (pas)' There are two possible analyses of this process, one involving a special s/r deletion rule and the other, with a slight modification, being a consequence of three dimensional representation. Under the former analysis, the appropriate s/r deletion rule can not simply delete every s or r in the language that follows a consonant (*s/r > 0 / C]v ). The clusters s-s and n-r, for example, are very common even though the s/r deletion rule functions to delete the second segment in cases like (77c) and (77i). Neither can the appropri ate rule simply delete every s_ or r that follows a consonant-final stem (*s/r > 0 / C] ), since clusters such as n-r and n-s are regulari-ly found across ]N boundaries (e.g. ken]N+ ro]N > ken-ro 'steep path') (cf. 77i). In fact, s/r deletion only applies to the initial s or r of verbal suffixes, following consonant-final verb roots. A suitable s/r deletion rule, then, would be (78). Note that the rule also deletes the 52 skeletal point associated with the s or r. (78) s/r deletion X | > 0 / c]v__ s/r A sample analysis utilizing (78) is presented in (79). (79) hanas + sase + (ru) > hanas-ase-(ru) (77c) (a) level I, cycle I: extrametricality, syllabification $ $ / \ / \ 0 N 0 N 1 X X 1 X 1 X h 1 a n a s]v [ex} (b) level I, cycle II: derivation, loss of extrametricality $ $ $ $ / \ / \ / \ / \ 0 1 N 0 1 N 1 1 X 1 X 1 1 X l l X l 1 h 1 a 1 n 1 a N 0 l N 1 X 1 X 1 1 X 1 a 1 s 1 e] SJy Sao ^Jy At this point the formerly extrametrical s_ can be syllabified as a coda since it is governable by a following onset. (c) level I, cycle II: syllabification $ / \ $ / \ $ $ / \ / / \ / \ / \ 0 N 1 1 0 N 1 1 C 1 0 N 1 1 0 N 1 1 1 1 X X 1 1 1 1 X X 1 1 1 X 1 1 1 X X 1 1 1 1 X X I 1 1 1 h a 1 1 n a 1 s]v 1 1 s a 1 1 s e 53 The initial s of sase is then deleted by rule (78). (79) -(d) level I, cycle II: £ deletion $ / \ $ / \ $ / \ / / \ / \ 0 N 1 1 0 N C N 0 N 1 1 1 1 X X 1 1 X 1 X X 1 1 X | 1 1 X X 1 1 1 1 h a 1 n I 1 a s]v 1 a 1 1 s e] V Under this analysis, the deletion of the js seems unmotivated since it disrupts a stable syllable structure. Following this disruption, the former coda JB is reanalyzed as an onset for the isolated nucleus a. (e) level I, cycle II: resyllabification $ $ $ $ / \ / \ / \ / \ 0 N 0 N 1 1 0 N 0 N X X 1 1 X X X X X X 1 1 h a n a s]v a 1 1 s e] V The second possible analysis of s/r deletion posits that the ini tial s or r of verbal suffixes is lexically floating. The resultant analysis is more elegant than (79) since a rule that disrupts stable syllable structure is not required. (80) hanas + sase + (ru) > hanas-ase-(ru) (77c) (a) level I, cycle I: extrametricality, syllabification $ $ / \ / \ 0 1 N 1 0 1 N 1 1 X 1 X I 1 X 1 X h 1 a n a [ex] 54 0 1 N 1 0 1 N 1 1 X 1 X 1 X 1 X h a n a (80) (b) level I, cycle II: derivation, loss of extrametricality $ $ $ $ / \ / \ I / \ N 0 N I I I X XXX s]v s a s e]v Rules of syllabification proceed to syllabify (80b). In this case, the final s_ of hanas- can act as an onset for the isolated nucleus a of -sase. The resultant syllabification is completely unmarked and the rules are content. The floating s, then, is not required and simply drops out. (80) (c) level I, cycle II: resyllabification, deletion of floating segment $ $ $ $ / \ / \ / \ / \ 0 N 1 I X X s e]v This second, more elegant analysis is chosen in this thesis as the best alternative. 0 1 N 0 N | 0 1 N I 1 X X X 1 X 1 X 1 X h a n a. s" lv a VTII.3: Level II Ccnipouriding Revisited Returning to the examples in (73b), the syllabification of verb compounds whose second elements are vowel-initial is not as one might predict at first. Consider (81). 55 (81) yom + owar + (ru) > ycm-i-owar-(ru) (a) level II, cycle I: loss of extrametricality (cf. 76b) • $  $ / \ | / \ 0 N 1 I XXX I I I r [ex] Based on (80b) and (80c), the following syllabification and subsequent deletion of the floating _i might be expected. (82) * $ $ $ / \ / \ / \ N 0 1 N 1 X 1 1 X 1 1 X 1 I i]v[o 1 w 1 a 0 N 1 1 0 I N 1 1 X X 1 1 1 X X | 1 1 m]v[o w 1 a [ex] However, (82) does not occur. Instead, the floating _i is provided with a point, and resyllabification produces (81b). (81) (b) level II, cycle I: resyllabification $ $ $ $ / \ / \ I / \ 0 N N 0 1 N X 1 X X 1 1 1 X 1 X 1 1 m 1 1 i]v[o 1 w 1 a r... [ex] Apparently syllabification across a ][ boundary is prohibited, a con straint that prevents (82) above from occurring. 56 Notes for Chapter VIII: 1. Following a suggestion of Shaw (personal communication). 57 Chapter IX Level III The principal morphological process that takes place at level III is the affixation of verbal and adjectival inflections. Ten verbal and nine adjectival inflections are recognized by Bloch (1946). Verbal Adjectival indicative (Ind) -ru -i past indicative (PInd) -ta - 'katta non-past presumptive (Pve) -yo'o -karo 'o past presumptive (PPve) -taro 'o - 'kattaroo conditiona1 (Cnd) -ta 'ra - 'kattara alternative (Alt) -ta 'ri -'kattari provis iona1 (Pro) -re'ba - 'kereba infinitive (Inf) -0 -ku gerundive (Ger) -te - 'kute imperative (Imp) -0, -e, -ro As exemplified throughout this paper by the inflections -ru and -i, every verb and adjective must have one (and only one) inflectional suffix^ attached at the very end of the word. Inflections are always found outside the derivational suffixes of level I, and are never found inside level II compounds (e.g. *mi-ru]y+ kir-(ru)]v 'see-cut'). These facts indicate that inflections are added later than level II compound ing. In fact, phonological process associated with the affixation of these inflections show that they are found at a completely different level—level III. 58 IX. 1: Stem Formative Truncation First of all, it is evident from (84) that, in most cases, no stem formative i. or any other type of intervening i is inserted between consonant-final roots and consonant-initial inflections. (84) (a) tat]v+ ta]v > tat-ta 'stood' *tat-i-ta (b) kaw]v+ ta]y > kat-ia 'bought' *kaw-i-ta (c) hur]v+ ta]v > hut-ta 'rained' *hur-i-ta Clearly the stem fornv-.tive that persisted through levels I and II is somehow truncated by the time it gets to level III. In some sense this truncation is reasonable. At early levels, the stem formative's func tion was derivational, to ensure that all verb roots were vowel-final. Since all derivational morphology is completed by the end of level II in Japanese, there is no need for stem formatives beyond that point. The complete derivation of (84c), then, is presented in (85). (85) hut + ta > hut-ta (84c) (a) level I: extrametricality, syllabification, stem formative $ / \ 0 N 1 I XXX I I I h u t]v i]v [ex] 59 (b) level III: stem formative truncation, inflection, loss of extrametricality $ $ / \ / \ ON ON X X X X X I I I I I h u t]v t a]v (c) level III: syllabification, coda government principle, obligatory contour principle $ / \ / \ $ / / \ / \ 0 N C 0 N 1 I I ! I X X X X X II \ / I h u t a]v IX.2: Infinitives/Connectives The Japanese language frequently uses i-final verbal constructions. These are known variously as infinitives (Martin, 1975) or connectives. (86) (a) yom]v+[ > yam-i 'read (inf).' (b) sin]v+[ > sin-i 'die (inf) ' (c) oyog]y+[ > oyog-i 'swim (inf) ' Traditional grammarians identify this final i as a stem formative, presuming it to be of the same origin as any other stem formative^. However, it has been shown in the previous section that stem formatives are truncated at the beginning of level III. This means that the final i of the examples in (86) cannot possibly be a stem formative, since it persists through level III. Furthermore, unlike a stem formative, the final i_ of the infinitive must be introduced complete with a skeletal 60 point. Consider (87). (87a) represents the first step in the deriva tion of yom-i read (inf)'. The target structure is represented in (87b). [87) yom +• i > yom-i (86a) (a) level I: extrametricality, syllabification, stem formative $ / \ 0 N XXX y o m]v i]v [ex] (b) level III: $ $ / \ / \ 0 N 0 N X X X X y o m]v i]v [ex] If the floating stem formative in (87a) were the same i_ present in (87b), then it would be difficult to motivate the target syllabification of (87b). In the first place, the i does not start out with a skeletal point for syllabification rules to work on and, secondly, the m is extrametrical and, hence, invisible to the syllabification rules. Instead, it is more plausible to truncate the stem formative at the beginning of level III (87c) and then reintroduce an i_ (with a skeletal point) as an infinitival inflection (87d). I conclude, there fore, that the final i_ of yom-i is actually an infinitival inflection rather than a stem formative. A major difference between the two i/s, 61 then, is that the inflectional i is accompanied by a skeletal point, whereas the stem formative i_ is not. (87) (c) level III: stem formative truncation $ / \ 0 1 N 1 1 X 1 1 X X 1 y o ii] [ex] (d) level III: inflection, loss of extrametricality, resyllabification $ $ / \ / \ 0 N 0 N 1 I I I X X X X y o m]v i]v [ex] IX.3: Real Epenthesis An true case of i_ epenthesis at level III occurs between s-final roots and consonant-initial inflections. (88) hanas]v+ ta]v > hanas-i-ta 'speak (PInd)' os]v+ ta]v > os-i-ta 'push (PInd)' mas]v+ tara]v > mas-i-tara 'increase (PInd) ' Unlike the derivational stem formative -i_ and the infinitival inflec tional i. discussed above, this -i has no morphological function whatso ever. It is strictly phonological, occurring to break up the impermis sible sequence s-t. Consequently, I propose to introduce the -i by the rule formulated in (89). 62 (89) epenthesis (level III) X [+cor] 0 > | / [+str] C i IX. 4: s/r Deletion An s/r deletion phenomenon has already been identified at level I (section VTII.2). The same floating segment analysis is required at level III to delete the r of r-initial inflections following consonant-final roots. (90) suwar]y+ ru]y > suwar-u 'sit' hanas]v+ £ulv > hanas-u 'speak' asob]v+ reba]v > asob-eba 'play (pro)' nom]v+ reba]v > nom-eba 'drink (pro)' (91) kaw + ru > kaw-u 'buy' (a) level III: inflection, loss of extrametricality $ $ /\ I ON N I I I XXX X k a w]v r u]v (b) level III: resyllabification, deletion of floating segment $ $ / \ / \ 0 N 0 N 1 I I I X X X X k a w]v u]v 63 IX.5: Accent Each Japanese word exhibits either one accent or none at all, regardless of whether that word is a noun, verb, compound, or stem with a long string of suffixes. Situations frequently arise where more than one accent may be assigned to a word by accent rules or by principles of lexical assignment during its derivation. In these cases, there are basic principles which decide which accent will predominate and become the single accent of the word. It so happens that the accent predomin ation principle of level I is the same as that of level II, while that of level III is different. This provides additional evidence for the autonomous status of level III. At level I a verb root will receive an accent on its penultimate mora by the V/A accent rule on the first cycle (e.g. tabe- 'eat'). When a derivational suffix is added on a subsequent cycle, a second ac cent is added (e.g. ta'be-sa'se- 'eat (caus)'). In such a case it is always the second (or rightmost) accent that predominates (i.e. tabe- sa'se-), while the initial accent is eliminated. No matter how long the concatenation of suffixes is, the Right Accent Predomination Principle (RAPP) still holds (e.g. tabe-sase-rare-ta-ga r-). At level II, conflicting patterns of accentuation can arise in several situations. For example, the accent of Sino-Japanese compounds of level I is lexically determined. When 'ka is added to such com pounds, the resultant combination may contain two accents. Once again, RAPP deletes the leftmost accent and preserves the right. 64 (92) ge'n + dai + 'ka > gen-da'i-ka 'modernization' /ge'n + dai + 'ka/ level I moncmorphemic compd ge'n + dai level II compounding RAPP ge'n-da'i + ka gen-da 'i-ka gen-da 'i-ka The compound noun accent rule of level II is complex and a com plete discussion of it is beyond the scope of this paper (see McCawley (1977)). Suffice it to say that regardless of the accentuation of the individual elements, the accent of the compound, if present, is always borne by the rightmost member. (93) no'ogyoo]N+ kumiai]N > noogyoo-ku'miai 'agricultural union' tu'uka]N+ ryo'ori'JN > tuuka-ryo'ori 'Chinese cooking' iso'ppu]N+ monoga'tari]^ > isoppu-monoga'tari 'Aesop's fables' By contrast, at level III it is the leftmost accent that survives accentual conflicts. Several of the verbal and adjectival inflections in (83) contain lexical accents which surface when they are attached to unaccented stems. However, when these inflections are added to accented stems, the accent of the stem is preserved by the Left Accent Predomina tion Principle (LAPP) and the accent of the inflection is lost. (94) (a) unaccentable stems ake + ta'ra > ake-ta'ra susum + re'ba > susum-e'ba kura + 'katta > kura'-katta 'open (end) ' 'advance (pro)' 'was dark' (b) accentable stems ta'be + ta'ra > ta'be-tara 'eat (end)' tano'm + re'ba > tano'm-eba 'ask for (pro)' ta'ka + 'katta > ta'ka-katta 'was high' 65 On the topic of accent, a rule of V/A accent adjustment will also be needed to account for the "accent attraction" phenomenon of the indi cative adjectival inflection -i_ and the r-initial verbal inflections following vowel-final stems. This readjustment rule will draw the ac cent from its basic stem-penultimate mora position and reassign it to the stem-final mora position. (95) (a) adjectives ta'ka + ku > ta'ka-ku 'is high (inf)' ta'ka + i > taka'-i 'is high' (*ta 'ka-i) (b) verbs ta'be + ta > ta'be-ta 'eat (PInd)' ta'be + ru > tabe'-ru 'eat' ta'be + re'ba > tabe'-reba 'eat (pro)' (*ta 'be-ru) (*ta'be-reba) 66 Notes for Chapter IX: 1. Many of the adjectival inflections are clearly bimorphemic, being composed of an adjectival element followed by the corresponding ver bal inflection from column 1. The most common adjectival element is 'kar which is found in -"kat-ta, -kar-ob, -'kat-taroo, -Tcat-tara, -and -"kat-tari. The coda government principle transforms the r of 'kar to t in most of the examples. The adjectival inflections -i, - 'ker-eba, -ku, and - 'ku-te are exceptional. 2. Since traditional grammarians consider the final i to be a stem formative and not an inflection, and since each verb must end with an inflection, traditional grammarians have been forced to invoke a 0 inflectional suffix for infinitives (see (83)). 67 Chapter X Conclusion The following morphological and phonological processes have been identified at the three levels within the Japanese lexicon. (96) morphology level I monomorphemic compounding stem formative derivation: -sase -rare -ta -rasi -gar -kata phonology extrametrica1ity high vowel syncopation V/A accent PAPP level II compounding RAPP level III stem formative truncation epenthesis inflection LAPP accent adjsutment In addition to the phonological processes identified above, a principle of coda government defined in (26) is assumed to hold through out the lexical component. The level ordered grammar outlined in (96) provides elegant solu tions to the two problems introduced in Chapter I involving i_ insertion and high vowel syncopation. With respect to i insertion, the desidera tive suffix -ta behaves differently from the homophonous perfective inflection because the two are found at different levels in the grammar: the former at level I and the latter at level III. More specifically, 68 the i which is introduced as a stem formative at level I and often appears in conjunction with the level I -ta is truncated at the begin ning of level III and, hence, is never found in conjunction with the level III -ta. In the same way, whether or not high vowel syncopation deletes the high vowel between two elements of a Sino-Japanese compound depends entirely on what level the compound was established at. If the compound was formed at level I, then high vowel syncopation will apply since the rule is found at level I. Compounds formed later than level I will never trigger high vowel syncopation. Perhaps most significantly, evidence from numerous morphological and phonological interactions has positioned each of the various proces ses of (96) at its appropriate level. The result is a well-motivated grammar that can be expected to shed light on issues beyond the scope of this thesis. 69 Bibliography Bloch, B. 1946. Studies in Collcquial Japanese I: Inflections. JAOS, 66, 97-109. Bloch, B. 1946. Studies in Colloquial Japanese III: Derivation of Inflected Words. JAOS, 66, 304-315. Chew, J.J. 1972. On Word Boundaries in Japanese. Journal Newsletter  of the Association of Teachers of Japanese, 2 Grignoh, A-M. 1985. Phonologie Lexicale Tri-Dimensionnelle du Japonais. Unpublished Ph.D. Thesis, Universite de Montreal. Haraguchi, S. 1977. The Tone Pattern of Japanese: an autosegmental  theory of tonoloqy. Tokyo: Kaitakusha Harris, J. 1983. Syllable Structure and Stress in Spanish: a non  linear analysis. Cambridge: M.I.T. Press. Ingria, R. 1980. Compensatory Lengthening as a Metrical Phenomenon. Linguistic Inquiry, 11, 465-495 Kageyama, T. 1982. Word Formation in Japanese. Lingua, 57, 215-258 Kaye, J.D. and J. Lowenstamm. 1983. Compensatory Lengthening in  Tiberian Hebrew: theoretical implications. MS. Kaye, J.D., J. Lowenstamm, and J.R. Vergnaud. (1985). Internal Struc ture of Phonological Elements: A Theory of Charm and Government. MS. Kempf, Z. 1972. Japanese Verb Compounds. Gengo Kenkyu, 62 Kiparsky, P. 1982. Lexical Morphology and Phonology. In van der Hulst, H. and N. Smith, (eds.). The Structure of Phonological  Representations (Part I). Kiparsky, P. 1982. How Are the Levels Linked? MS. Kiparsky, P. 1985. Some Consequences of Lexical Phonology. In Ewen, C.J. (ed.). Phonology Yearbook 2. Cambridge: University Press. Koo, J.H. 1974. Reinterpretation of the Desiderative -tai and its Related Morphophonemics. Journal Newsletter of the Association  of Teachers of Japanese, 9. Leben, W.R. 1980. A Metrical Analysis of Length. Linguistic Inquiry, 11,497-509. McCarthy, J. 1981. A Prosodic Theory of Nonconcatenative Morphology. Linguistic Inquiry, 12, 373-418. 70 McCawley, J. 1968. The Phonological Component of a Grammar of  Japanese. The Hague: Mouton. McCawley, J. 1977. Accent in Japanese. Studies in Stress and Accent. Hyman, L. (ed.) Los Angeles: University of Southern California. Maeda, S. 1979. Boundaries in Japanese Phonology. Gengo Kenkyu, 75 Marantz, A. 1982. Re Reduplication. Linguistic Inquiry, 13, 435 Martin, S.E. 1952. Morphophonemics of Standard Colloquial Japanese. Language Dissertation, no. 47. Baltimore: Linguistics Society of America. Martin, S.E. 1975. A Reference Grammar of Japanese. New Haven: Yale University Press. Mohanan, K.P. 1981. Lexical Phonology. Ph.D. Dissertation, M.I.T. Parker, C.K. 1939. A Dictionary of Japanese Compound Verbs. Tokyo: Maruzen. Rubach, J. 1984. Segmental Rules of English and Cyclic Phonology. Language, 60, 21-54. Shaw, P.A. 1985. Modularisation and Substantive Constraints in Dakota Lexical Phonology. In Ewen C.J. (ed.). Phonology Yearbook 2. Cambridge: University Press. Takenobu, Y. (ed.) 1940. Kenkyusha's New Japanese-English Dictionary. Kenkyusha: Tokyo. van der Hulst, H. and N. Smith, (eds.) The Structure of Phonological  Representations (Part I). An Overview of Autosegmental and Metrical Phonology, pp. 1-46 Yip, M. 1982. Reduplication and C-V Skeleta in Chinese Secret Lang uages. Linguistic Inquiry, 13, 637-661. Zubizaretta, M.L. 1979. A Metrical Account of the Tone Pattern of Japanese: the interaction of harmony and accent. M.I.T. MS. 71 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0096517/manifest

Comment

Related Items