ON THE STRUCTURE OF SYNCRETISM IN ROMANIAN CONJUGATION RONALD F. FELDSTEIN Indiana University 0. Introduction Romanian conjugation displays several cases of syncretism, in which two paradigmatic slots share the same grammatical desinence.1 This syncretism occurs in both the present and imperfect tenses of the indicative, as well as in the subjunctive. In the present tense, syncretism occurs in the 1sg/3pl, the 3sg/3pl, and the 1sg/2sg forms; in the imperfect, the 1sg/1pl forms are always syncretic; number syncretism is always found in the 3sg/3pl of the subjunctive. I will claim that all such instances of syncretism should be divided into two categories, based on whether they are phonologically conditioned or not. When syncretism is phonologically conditioned, it will only apply to a specific phonological subset of verbs, while the nonphonological type applies across the board. In Romanian, the present-tense varieties of syncretism can be explained by phonological conditioning, once the appropriate underlying morphophonemic constructs and rules are established. On the other hand, the syncretisms of the imperfect and subjunctive are not phonologically conditioned and, as such, apply to every verb without exception. Two recent publications attempt to analyze the syncretisms of Romanian conjugation: sections of a book by Stump (2001:213-215) and a paper by Bobaljik (2002:65-66). This chapter will show that both authors operate with structurally inadequate models of Romanian conjugation and syncretism, which leads them to make analytical errors. Their errors are mainly due to the fact that they do not distinguish phonologically conditioned, or variable syncretisms, from invariant syncretisms that have nothing to do with a 1 This chapter represents Romanian sounds by means of orthographic symbols. Letters that should be specially noted are ă (equivalent to the mid central vowel schwa, i.e., []), and î, which represents the high central vowel []. The high vowels i and u, in postvocalic position (e.g., dai, dau) represent nonsyllabic [i] and [u]. The sequence ea represents the diphthong [ea]. The consonants ş, ţ represent [] and the affricate [ts], respectively. 178 RONALD F. FELDSTEIN particular phonological environment. These inaccuracies stem from the authors’ reliance on textbook notions of the various subtypes of Romanian syncretism, which are traditionally, but incorrectly, said to be determined by conjugation type. Since these are not really errors made by Stump and Bobaljik per se, but are caused by the inadequacies of the traditional treatment of Romanian conjugation and syncretism, it becomes clear that the traditional approach needs to be modified. This chapter will begin with a brief review of the basic facts about Romanian syncretism (section 1), where the traditional conjugations will be compared to the actual distribution of syncretic types. Next (section 2), a new segmentation and morphophonemic analysis of Romanian conjugation is offered, which resolves some of the difficult issues of Romanian syncretism. Phonologically conditioned syncretism will be treated as underlyingly non- syncretic; that is, it should be considered to be a superficial, or surface, syncretism, while the nonphonological variety remains as a deeper syncretism. Section 3 examines how Stump (2001) and Bobaljik (2002) rely on the textbook view of conjugation for their analyses of Romanian syncretism, and the flaws of this approach are pointed out. While some of these errors are due to not considering all the data, others stem from an inefficient segmentation of the verb stem and desinence. Section 4 looks at the issue of whether a surface perspective can offer any useful linguistic generalizations about Romanian syncretism. Since the work of Stump (2001) and Bobaljik (2002) is relevant to this chapter only insofar as it illustrates the inadequacy of the traditional textbook approach to present-tense syncretism, as based on conjugational type, rather than stem phonology, I will not specifically comment on the other main issue they discuss—whether these syncretisms should be treated as rules of referral2 (the notion that one paradigmatic cell is replaced with the other) or strictly as rules of impoverishment (the idea that both cells are neutralized with respect to a particular grammatical feature). Generally speaking, Stump opts for the use of rules of referral, while Bobaljik prefers to operate with the method of impoverishment. However, I would emphasize that one cannot even begin to make linguistic judgements about the merits of impoverishment versus referral without a more accurate notion of the structural units of Romanian syncretism. 2 Bobaljik (2002:66) maintains that impoverishment can handle all of the cases of Romanian syncretism and that rules of referral (introduced by Zwicky 1985:372) are too powerful and unrestricted a mechanism. Stump (2001) opts for unstipulated syncretism (equivalent to impoverishment) in some instances, but posits rules of referral in others. SYNCRETISM IN ROMANIAN CONJUGATION 179 1. Brief review of Romanian syncretism I maintain that present-tense syncretisms are a function of the phonological properties of the verb stem. The traditional system, as repeated by Stump (2001) and Bobaljik (2002), states that there are four basic conjugations, and that each conjugational type utilizes a different set of desinences (and syncretisms), unrelated to phonology. I claim that each of the traditional conjugations merely refers to a particular stem-final theme vowel, which can be considered a verbal formant. Furthermore, I propose that the underlying desinences are identical for all four of the so-called conjugations, and that the surface differences of conjugation are the result of the phonological interaction of the different stem-final themes with the unified set of desinences. In many cases, the traditional system of conjugation does a poor job of capturing what actually occurs, especially in the area of syncretism. Table 1 lists the four traditional conjugational types. Type I has the theme vowel -a and usually has 3sg/3pl syncretism; types II, III, and IV have the theme vowels -e, -ea, -i, (and -î3) and most frequently have 1sg/3pl syncretism. Table 2 presents present-tense paradigms for these and other verbs. Conj. Stem type Examples (infinitive) I -a invita “invite,” tăia “cut,” sufla “breathe,” da “give” II -ea tăcea “be silent,” umplea “fill (variant),” bea “drink” III -e bate “beat,” umple “fill” IV -i sări “jump,” coborî, “descend,” sui “climb,” suferi “suffer,” şti “know” Table 1: Traditional Romanian conjugations (Graur 1966:246) Infinitive Conj. type 1sg 2sg 3sg 1pl 2pl 3pl invita I invít invíţi invítă invitắm invitáţi invítă tăia I tái tái táie tăiém tăiáţi táie sufla I súflu súfli súflă suflắm sufláţi súflă da I dáu dá́i dắ dắm dáţi dáu tăcea II tác táci táce tăcém tăcéţi tác umplea II úmplu úmpli úmple úmplém úmpléţi úmplu bea II beáu béi beá́ bém bé́ţi beáu báte III bát báţi báte bátem báteţi bát úmple III úmplu úmpli úmple úmplem úmpleţi úmplu sắrí IV sár sári sáre sărím săríţi sár coborî IV cobór cobóri coboáră coborîmi coborîţi coboáră suí IV súi súi súie suím suíţi súie suferi IV súfăr súferi súferă suferím suferíţi súferă ştí IV ştíu ştíi ştíe ştím ştíţi ştíu Table 2: Sample conjugations of verbs in Table 1, with syncretic cells marked in bold 3 Traditional grammar places theme vowels –i and –î in conjugation IV. 180 RONALD F. FELDSTEIN Table 2 shows that all of the verbs of a given conjugation type (i.e., I, II, III, IV) do not necessarily share the same pattern of syncretism. The following two reasons show why it is therefore wrong to maintain the traditional conjugation types when speaking of syncretism: (i) Verbs of the same traditional conjugation type (i.e., with the same theme vowel) can have different patterns of syncretism. For example, invita “invite” has syncretism of the 3sg/3pl type. However, tăia “cut” has both the 1sg/2sg and 3sg/3pl types, while da “give” has the 1sg/3pl type, in spite of the fact that all of these stems end in the theme vowel –a. Although these verbs are ostensibly all of the same type, their different types of syncretism are due to the fact that the theme is not monosyllabic in the first two verbs, but is monosyllabic in the third (da); the monosyllabic stem causes the theme vowel to be stressed in all present forms, and conditions syncretism of the 1sg/3pl type, in spite of the –a theme. Furthermore, the second verb (tăia) has [i] (‘yod’) as its stem-final consonant, which causes an additional 1sg/2sg syncretism. In other words, phonology, rather than conjugational class, causes the specific pattern of present-tense syncretism. (ii) Verbs of different traditional conjugations, often assumed to have different syncretic patterns, can actually have the same type of syncretism, if certain phonological conditions are met. For example, type I invita and type IV sui and coborî all share syncretism of the 3sg/3pl type. Phonologically, this is conditioned by the fact that all have back vowels either as the theme or immediately preceding the theme vowel, in spite of the fact that they do not belong to the same traditional conjugational classes. In other words, the patterns of present-tense syncretism are predictable on the basis of the phonological shape of the stem, rather than the traditional conjugation class, so it is incorrect to identify particular syncretic types with the traditional conjugations. Table 3 demonstrates that each major type of present-tense syncretism can be correlated with a variety of different traditional conjugation types (i.e., theme-vowel classes). For example, Table 3 shows that syncretism of the 1sg/3pl type can be manifested by verbs of all four conjugation types. Furthermore, each concrete manifestation of syncretism can be correlated to specific phonological properties of the stem. As illustrated, the 1sg/3pl syncretic form can end in the nonsyllabic glide [u], vocalic [u], or a consonant. In the first instance, the stem is monosyllabic and can have a theme vowel that cuts across the broad range of traditional conjugations I, II, and IV. These cases are not isolated exceptions, but defined phonological types. SYNCRETISM IN ROMANIAN CONJUGATION 181 Syncretic Phonological manifestation of syncretisms cells 1sg/3pl 1. Syncretic form ends in non-syllabic glide -u (theme vowels –a, -ea, -i): present Monosyllabic stems, e.g., 1sg/3pl dau, beau, ştiu (infin. da/bea,şti) 2. Syncretic form ends in the vowel –u (theme vowels –ea, -e): Front-vowel theme preceded by [obstruent + liquid] cluster, e.g., 1sg/3pl umplu (infin. umplea ~ umple) 3. Syncretic form ends in zero (followed by pretheme consonant; theme vowels –ea, -e, -i): Non-monosyllabic stems, e.g., 1sg/3pl tac, bat, sar (infin. tăcea/bate/sări) 3sg/3pl 1. Syncretic form ends in –ă (theme vowels –a, -i, -î): present a. Non-monosyllabic stem in non-front vowel -a or -î, preceded by nonalatal, e.g., 3sg/3pl invită, coboară (infin. invita, coborî) b. Polysyllabic stems in theme vowel -i, where present tense stress falls two syllables before the theme-vowel. e.g., 3sg/3pl spríjină “support,” súferă “suffer” (infin. sprijini/suferi) 2. Syncretic form ends in –e (theme vowels –a, -i): a. Non-monosyllabic stem ending in palatal glide plus -a (-ia), e.g., 3sg/3pl taie “cut” (infin. tăia) b. Stems ending in a two-vowel sequence, the first of which is non-front, e.g., 3sg/3pl suie “climb” (infin. sui) 1sg/2sg Syncretic form ends in palatal glide (-i): present Non-monosyllabic stem ending in palatal glide plus -a (-ia), e.g., 1sg/2sg tai “cut” (infin. tăia) 1sg/1pl All verbs imperfect 3sg/3pl All verbs subjunctive Table 3: Syncretic types in Romanian conjugation For example, several monosyllabic stems, whose only vowel is the stem-final theme (e.g., da “give,” sta “stay,” bea “drink,” vrea “want,” şti “know,” etc.) all share the same syncretic pattern of 1sg/3pl, realized by the glide [u] (spelled u, e.g., 1sg/3pl dáu, stáu, beáu, vreáu, ştíu). The phonological reason is that the underlying -u, found in both the 1sg and 3pl, is changed to the glide when immediately following a stressed vowel, which is inevitably the case in monosyllabic stems. Therefore, the conditioning factor for this syncretic type has nothing to do with the traditional conjugation classes, but is directly related to the phonological feature of monosyllabicity. Thus, even a small sample of verbs should be sufficient to convince the observer of the direct causal role of stem phonology in determining the resulting syncretism. 182 RONALD F. FELDSTEIN Likewise, the syncretic pattern of 3sg/3pl is manifested by a final mid vowel: -ă (phonetically, []) or –e. As in the case of 1sg/3pl syncretism, a wide variety of conjugational types can manifest this syncretism, all of which can be defined on the basis of the specific phonological properties of the stem. Thus, we see that non-monosyllabic stems with a basic non-front theme-vowel (-a or –î, phonetically []) regularly have 3sg/3pl syncretism, but stems with basic front theme vowels only have this type of syncretism in narrowly defined conditions, such as when the front theme vowel is immediately preceded by a back vowel (e.g., infinitive sui “climb,” 3sg/3pl suie) or is not contiguous to the present-tense stress (e.g., infinitive sprijini “support,” 3sg/3pl spríjină). Table 3 also displays the difference between the phonologically conditioned types of syncretism found in the present tense and the non- phonological types found outside the present tense. For example, in the imperfect, first person is syncretic for number and in the subjunctive third person has number syncretism. However, these syncretisms are not based on phonology and apply to all verbs. By definition, if a syncretic type pervades an entire grammatical category, such as the first-person syncretism of the Romanian imperfect, there will be no phonological subcategories that condition the presence or absence of the syncretism. Therefore, it is obvious that the number syncretisms found in both the imperfect and subjunctive are qualitatively different from those of the present tense, where phonological conditioning is the rule. 2. A new segmentation and rule system 2.1 Segmentation of present-tense and imperfect desinences In this section, a new segmentation of Romanian verbs is proposed, with a unified set of grammatical endings. The variable realizations and syncretisms among the various verb stems are not caused by the fact that these stems lexically belong to a particular conjugation, but by the phonological interaction of the stem-final (theme) vowel and the ending. Each grammatical ending of the conjugational system is really a complex of three morphemes, representing the grammatical categories of tense, number, and person, in that order. Conjugation can be defined as the cyclical combination of the stem-final segment with each of the three components of the grammatical ending. The rules are morphophonemic, rather than strictly phonological, in that many of them only occur within the process of conjugation. There is no absolute neutralization of any of the posited grammatical morphemes. Each of the assumed morphemes surfaces in its basic form in at least some environments. In cases where certain morphemes (e.g., word-final -u) do not surface, this SYNCRETISM IN ROMANIAN CONJUGATION 183 occurs due to general phonological rules that exclude such occurrences.4 My notation will show the basic stem followed by the ‘+’ symbol, for example, invita+; components of the grammatical desinence are separated by hyphens. Zero morphemes are rendered with the symbol ‘Ø,’ and word-final position is indicated with the symbol ‘#.’ Each zero morpheme is opposed by at least one instance of a non-zero. Tense Number Person Tense Number Person 1sg Ø Ø -u 1pl Ø -m -u 2sg Ø Ø -i 2pl Ø -t -i 3sg Ø Ø Ø 3pl Ø -u Ø Table 4: System of present-tense endings Tense Number Person Tense Number Person 1sg -à -m -u 1pl -à -m -u 2sg -à Ø -i 2pl -à -t -i 3sg -à Ø Ø 3pl -à -u Ø Table 5: System of imperfect-tense endings The system of present- and imperfect-tense endings can be found in Tables 4 and 5. The middle morpheme position, that of number, significantly differs from the others, in that there is no basic phonological representation for the plural, although the singular can be treated as a consistent zero. In any case, it is assumed that some prior mechanism will make these morphemes available for conjugation in the way they are represented in Tables 4 and 5. The present- tense morpheme is a zero, and is opposed to the stressed vowel morpheme -á of the imperfect.5 The -á does, in fact, surface in all forms of the imperfect. The process of commutation suggests a zero for the present, since certain 4 One might question the assumption of a final -u for first person in the 1pl form, since it can never surface in this form for phonological reasons, that is, it can only be preceded by nasal - m, while a preceding [consonant + liquid] cluster is required for final -u to surface. However, both the parallelism with the 1sg and the possibility of an underlying final -u argue in favor of its underlying existence. 5 I posit stressed -ú for the simple perfect. Due to space considerations and the fact that there is no syncretism in the simple perfect, the topic is not discussed in this chapter. 184 RONALD F. FELDSTEIN imperfect desinences are longer than the corresponding present-tense endings by one segment (e.g., bateam “beat ” vs. batem “beat ”). The 1SG/PL.IMPF 1PL.PRESENT consonantal desinences, which represent number in the first and second persons, regularly surface as such, except for the fact that 2pl basic -t is palatalized to -ţ, due to its position before a word-final -i. In fact, unstressed word-final -u and -i are subject to the most variation. They survive intact only when preceded by a [consonant + liquid] cluster (e.g., aflu, afli); when preceded by a single consonant or another type of consonant cluster, the -u is deleted (e.g., 1sg invit), while the -i is reduced to a nonsyllabic glide (i), which often palatalizes the preceding consonant (e.g., 2sg inviţi). 2.2 Rules for vowel → mid and vowel deletion In the imperfect tense, first-person syncretism is not phonologically conditioned; I assume that it is grammatically assigned and appears as the input to the phonological rules, as shown in the identical first-person slots of Table 5. By contrast, the syncretic cells of the present tense come into existence due to the application of morphophonemic rules. Two specific types of morpho- phonemic rules are of particular importance for the generation of present-tense forms; their function will be the next major topic of discussion. Theme vowels all surface as mid at certain points in the present tense, even if they are high or low vowels in the basic stem. This suggests that there is a rule that lowers high vowels to mid and raises low vowels to mid. This is most obvious in the 3sg form, where the corresponding infinitives have high and low stressed vowels (cf. invitá, sărí), yet these vowels all surface in the present as unstressed vowels that have changed to mid (cf. 3sg invítă, sáre). Therefore, the first major sort of rule will require non-mid vowels to change to mid in particular environments. When the basic (unstressed) theme vowels totally fail to surface in certain present-tense forms (such as 1sg), even as changed to mid, we can observe that the following underlying desinential vowel surfaces instead, leading to the conclusion that a sequence of two underlying unstressed vowels experiences the deletion of one of these vowels. Since high and low stressed vowels (found in monosyllabic stems) do not undergo deletion in the 1sg, but instead cause the -u to become a glide (cf. ştíu, dáu), I assume that the deleted theme vowels experienced deletion before -u because they had first become unstressed mid vowels. In the 1sg, the vowel that survives the deletion (final - u#) is itself subject to later deletion, unless a [consonant + liquid] sequence precedes it (e.g., invit, but suflu). Therefore, the surface forms do not obviously demonstrate that a vowel sequence and a deletion have occurred. Yet, if we start from basic stems invita+ and sufla+, deletion in both cases is the only SYNCRETISM IN ROMANIAN CONJUGATION 185 way to generate invit and suflu. In the 1sg form, all of the verb classes show a similar type of deletion. However, verbs are dramatically different in the 3pl and that is the key to solving the riddle of how syncretism is generated. Since front-vowel themes (i.e., -i and -e) experience the same deletion in 3pl and 1sg, those two forms are syncretic. The results of the commutation process (shown in Table 4) indicate that the underlying grammatical endings of the 1sg and 3pl differ in that 1sg has an -u desinence in the person slot (i.e., in word-final position), while 3pl has an -u desinence of number, which occupies the desinential medial slot, not directly on the word-final boundary. Since front-vowel themes (-e, -i) both experience the identical deletion in these two instances (1sg and 3pl, e.g., bat, sar), I assume that these themes undergo the same rules for being deleted before -u in both positions. However, the back- vowel themes (both -a and -î) experience deletion only in the 1sg (e.g., invit, cobor), but not in the 3pl (invită, coboară), where they surface as nondeleted mid vowels. This indicates that the high front theme vowel (-i) generalizes mid-vowel height before any unstressed desinential vowel, regardless of whether it is in word-final position, but that the high and low non-front theme vowels (-a, -î) do not undergo the change to mid-vowel height before a nonfinal -u (such as in the 3pl), but do so only at the point in the cycle when they reach the word-final boundary. Therefore, due to rule ordering, if a basic -a theme has not yet been raised to mid, it will not be deleted before a following -u desinence; rather, the -u desinence itself will be deleted and only later will the -a reach the word-final position and be raised to mid, surfacing as -ă (cf. 3pl invită). If, on the other hand, an -i theme gets lowered to mid even before being combined with desinential medial -u, the sequence -e-u- will ensue and the first vowel will be subject to deletion, first yielding a form such as bat-u#, which later loses the final -u and surfaces as 3pl bat, syncretic with 1sg. Thus, at each cycle of conjugation (stem + tense, stem + number, and stem + person), there will be rules for changing theme vowels to mid-vowel height and then for deleting vowels in the sequence of theme vowel plus vocalic desinence, as follows: (i) Any unstressed front-vowel theme (regardless of frontness or backness) becomes mid before an unstressed desinential vowel. This rule can take the form (1), where the ‘+’ symbol refers to the stem-desinence boundary: