Łedu

Phonology (Łłinjxäkoll/Ļiɲ̈äkol)

---

The Łedu phonology is vast, consonant-heavy, an arduous mouthful for most non-native speakers, and well-integrated with the morphology, but ultimately yields a relatively simple analysis. Despite – or perhaps as a function of – the language’s massive inventory of contrastive phonemes, instances of allophony are rather infrequent. Although it does adhere to strict consonantal voicing harmony in clusters, the morphology itself presents few junctions where consonants would come into contact such that assimilation occurs. Conversely, vowel adjacency occurs exclusively at morpheme boundaries (never within atomic morphemes) – but some manner of slight constriction always intervenes based on the preceding vowel’s roundedness or nasality features, so pure vowel contact happens ideolectally at most. In some morphological positions, instances of /ə/ are also typically dropped, such as in word-final postpositional endings or coordinative suffixes, where their pronunciation is considered an archaism. Otherwise, vowels are never dropped, and reduced only in length or pitch, when unstressed. Stress patterns hold that the first syllable of any word’s root morpheme(s) receives primary stress, while pitch ascends prior to this syllable, and descends subsequently. Sentential stress and prosody are loose, but pattern roughly according to the elements one wishes to focus or emphasize (including in sentences that don’t explicitly focus any element).

Syllable Structure

The most minimal Łedu syllable consists of V, and the hypothetical most maximal permitted by the phonology, CCCCCCVCCCCCCCCCC (such a hypothetical standalone syllable could be, for example, skfmlwahhrhllmbbzxqłhd – although no such syllable is even remotely attested). As this illustrates, Łedu is a language that permits, but does not require, ample consonant clusters. Basal syllables consist of a vowel nucleus, then prioritize an onset, then lastly, a coda. No more than one vowel, however, is permitted to exist in the nucleus aside from the case of diphthongs, which are counted as a single vowel.

In such a case that a single consonant exists between two vowels, it will always be considered the second syllable’s onset rather than the first syllable’s coda. But, while Łedu does accoustically/descriptively obey the Maximal Onset Principle, in such a case that two consonants exist between two vowels, the first will be considered the first syllable’s coda for purposes of morphology/compounding, and the second, the second syllable’s onset. In such a case that three or more consonants exist between two vowels, the sonority hierarchy detailed below (under Sonority and Syllable-Internal Positioning) will determine whether they fall into the former syllable’s coda or the latter syllable’s onset for those purposes – and if one or more consonants fall into such circumstances that would allow them in either position, they will alternate between joining the latter onset and the former coda, beginning with said onset. Note that [h] is incapable of appearing in codas at all, so it will always be considered part of the onset.

Additionally, take note that /l/ bears the allophone [ɫ] in coda position, but remains [l] in onset position. This difference is non-contrastive, but is nonetheless represented in the native orthography and the Latinate C-Łedu orthography, using a doubled <l>, <ll>, in coda position. This is not represented in the Latinate N-Łedu orthography, which uses <l> in both onset and coda.

Table of Consonants

* [q], [ʀ̥], and [χ] are non-contrastive allophones of /ɢ/, /ʀ/, and /ʁ/ respectively, when adjacent to other voiceless consonants. The difference is unmarked in the orthography.

**[ɸ] and [β] are allophones of the bilabial trills [ʙ̥] and [ʙ], respectively, in coda position.

*** The realization of [ʔ] varies from speaker to speaker, conditioned in certain morphological circumstances that are often marked in the orthography with <’>. Some speakers insert it, some do not.

Consonant Voicing Harmony

The regiment of consonants in Łedu exhibits voice harmony and assimilation where possible. When consonants that partake in contrastive voiced-voiceless pairs (stops, fricatives, affricates, lateral fricatives, and trills) appear adjacent to each other, they will always align in terms of voice features. In the case that two such consonants meet at a morpheme boundary and possess incongruent voice features, one will assimilate to the other, per a set of rules detailed below. When a consonant from the aforementioned groups appears adjacent to a consonant that does not partake in contrastive voicing (these include nasals, approximants, lateral approximants, and clicks, as well as [h] and [ʔ]), it does not change its voice feature regardless of what voicing the other consonant exhibits – unless it is uvular, which always assimilate.

To determine voicing assimilation at the junction of two or more conflicting, contrastive-voicing consonants at a morpheme boundary, obey the following set of rules:

1. Voiceless Assimilation to Voiced: If both consonants have the same manner of articulation, assimilate the voiceless consonant such that both are voiced. Eg. /(_V)k+ɟ(_V)/ → [(_V)g.ɟ(_V)]
2. Sonorous Assimilation to Occlusive Voicing: If the consonants differ in their manner of articulation, consult the sonority hierarchy in the section below, and assimilate the more sonorous consonant such that it has the same voice feature as the less sonorous. Eg. /(_V)r+s(_V)/ → [(_V)r̥.s(_V)]
3. Uvular Assimilation to Other Voicing: If one of the two consonants is uvular, assimilate it to the voice feature of the other consonant, regardless of what that other consonant is (unless the circumstances of rule 3 are met, in which case, proceed accordingly). If both consonants are uvular, proceed according to rule 2. Eg. /(_V)ɢ+r̥(_V)/ → [(_V)q.r̥(_V)]
4. Sonorous Assimilation to Occlusive-Pincered Uvular Voicing: If one of the consonants is uvular, and it is already adjacent to a consonant with a particular voicing feature, decide whether the uvular consonant (and potentially its existing neighbor) or its newly adjacent consonant assimilate, by weighing the newly adjacent consonant against the hitherto-adjacent consonant on the sonority grounds of rule 2. Eg. /(_V)sq+b(_V)/ → [(_V)zɢ.b(_V)]
5. Identical Consonant Absorption: If both consonants are exactly the same, the consonant in onset position absorbs the consonant in coda position. It does not undergo any changes, besides reducing the instances of said consonant at the morpheme boundary from two to one. Eg. /(_V)ɬt+t(_V)/ → [(_V)ɬ.t(_V)]
6. [h]-Mutation: If the former consonant is a fricative, lateral fricative, click, trill, or non-lateral approximant, and the latter consonant is [h], mutate it to either [ʀ̥] or [ʀ], matching the voicing of the final consonant in the coda that precedes it. Eg. /(_V)z+h(_V)/ → [(_V)z.ʀ(_V)]

Sonority and Syllable-Internal Positioning

Like most languages that include consonant clusters, Łedu’s phonology operates on the basis of an implicit sonority hierarchy, which determines which consonants may precede or follow others in a cluster. The order occurs as follows (where P = stop, F = fricative, N = nasal, Ƀ = bilabial trill, R = other trill, X = click, L = lateral approximant, Ł = lateral fricative, W = approximant, . = syllable boundary, and (_V) = nuclear vowel). A numbered asterisk before a particular node denotes an exception or caveat that will be detailed below.

Onset: .^1*(N)P→(^2*(N)F/^3*Ł)→^1*P→^4*(N)Ƀ→(^2*F/^3*Ł)→^5*N→(^6*R/^7*X)→^8*L→W→ (_^V)

Coda: (_^V)→^6*R→^9*W→L→^5*N→^4*Ƀ→(^2*F/Ł)→^7*X→P→P→(F/Ł)→P.

Exceptions/Caveats:

1. Stops: As indicated by the hovering ^(N) prior to the symbol P, stops in onset position may be prenasalized if they are not preceded by any other consonants. This manifests in the orthography by preceding such an oral stop with the nasal stop that corresponds to its place of articulation (<nh/ŋ> is used prior to <x̂>). Although an onset may contain two stops, they cannot be adjacent (*.PP) – a fricative must be present as a buffer (.P(F/Ł)P). Additionally, although .P(F/Ł)P is a possible onset, *.P(F/Ł)P(F/Ł) is not. Likewise, although .P(F/Ł)N is a possible onset, *.PN is not. Onsets such as, or containing, *.(_)PɃ and *.(_)PX are also impossible. Affricates may viably occur in whichever environments a stop that is followed by a fricative may also occur.
2. Fricatives: While .FP and .PF are both possible onsets, *.FPF is not. [h] cannot occur after any consonant in onset position, nor prior to stops, fricatives (besides [χ]), lateral fricatives, nasals, any trills (besides [r̥]), or clicks in onset position. [ʁ] behaves unlike other fricatives in onset position, as it is able to follow nasals if no other consonants precede them, violating the sonority hierarchy. Fricatives in onset position may rarely be prenasalized if not preceded by another consonant (depicted orthographically by appending the letter matching its place of articulation before it, as in mvrræhhäz- (black)), but very few such words exist. [h] cannot occur at all in coda position. [ʁ] cannot occur in codas containing approximants or trills, nor can it occur after any consonant in coda position – it must precede them.
3. Lateral Fricatives: While .ŁP and .PŁ are both possible onsets, *.ŁPŁ is not.
4. Bilabial Trills: Bilabial trills behave differently from other trills, meriting a unique spot in the hierarchy. They cannot follow stops in onset, but they can both precede and follow fricatives and lateral fricatives (although not at the same time, so *.(F/Ł)Ƀ(F/Ł) is impossible). They cannot precede nasals or clicks in onset, either, but can be prenasalized if no other consonants precede them (characterized by a preceding <m> orthographically). In coda position, they cannot precede clicks unless an intervening fricative is present (so while *BX. is impossible, Ƀ(F/Ł)X. is possible).
5. Nasals: No onset can include .(_)NX(_) in any capacity – nasals cannot be followed by clicks in onsets. Additionally, onset nasals may violate the sonority hierarchy in a key way: they may in fact precede [ʁ], allowing for an iteration of .NF when all other *.NFs are invalid (and *.PN[ʁ], *.FN[ʁ], and *.ŁN[ʁ] are all still invalid). In coda position, nasals cannot precede clicks unless an intervening fricative is present (so *NX. is impossible, as opposed to N(F/Ł)X.).
6. Trills: In onset position, [r̥/r] may not precede lateral approximants, but [ʀ̥/ʀ] may. In coda position, [r̥/r] may not precede an approximant, but [ʀ̥/ʀ] may. Neither may precede clicks.
7. Clicks: Clicks cannot follow stops, bilabial trills, or nasals in onset. In coda position, they cannot follow nasals nor bilabial trills without an intervening fricative present. When they precede stops in coda position, the only consonant that may follow is a fricative (so *XPP. is impossible, but XP(F/Ł). is, and another stop may theoretically follow that fricative). 
8. Lateral Approximants: Lateral approximants may not follow [r̥/r] in onset position.
9. Approximants: The only non-lateral approximant that can appear in coda position at all is [ɹ]. [j] and [w] cannot. 

When weighing sounds that appear multiple times in particular positions on the sonority hierarchy, like stops or fricatives, for purposes of allophony, one uses their most internal position for reference (therefore stops always overrule fricatives). The only exception is when weighing bilabial trills against fricatives, in which case fricatives always prevail and force trills to assimilate.

Table of Vowels

	Front	Center	Back
High	[i] <i> • [y] <y>		[u] <u> • [ũ] <ũ>
High-Mid	[ɪ] <ı>		[ʊ] <ü>
Mid-High	[e] <e> • [ø] <ø>		[o] <o>
Mid		[ə] <ę>
Mid-Low	[ɛ] <ė> • [œ] <ö> • [ɛ̃] <ẽ>		([ʌ])* <û> • [ɔ] <ô> • [ɔ̃] <õ>
Low-Mid	[æ̝] <æ>
Low	[æ] <ä>		[ɑ] <a> • [ɒ] <å> • [ɑ̃] <ã>

* [ʌ] is an allophone of /ə/ that is elicited in syllables with primary stress. All instances of [ə] in such syllables become [ʌ], but [ʌ] does not surface under any other circumstances. In all forms of the orthography, when [ʌ] occurs in a lexical root, it is marked (as <û> in any Latinate orthography), but in the few (largely pronominal) cases where it does not, it remains unmarked (as <ę> in any Latinate orthography). Native speakers tend to conceptualize [ʌ] and [ə] as separate sounds due to their stress-based and orthographic distinctions, despite the predictable allophonic pattern.

Table of Diphthongs

Dipthongs

[ɑi] <ai>

[ɑu] <au>

[oi] <oi>

Table of Intervocalic Constrictions

Constriction

[ʲ] <j> (palatal)

[ʷ] <w> (labial)

[ˠ] <x̂> (velar)

Intervocalic Constrictions

Between vowels that don’t have an intervening consonant, Łedu forms three constrictions that are perhaps too loose to be called proper consonants, but cannot be strictly attributed to the vowels themselves either. In IPA, they are denoted as in the table above. The palatal constriction [ʲ] will follow unrounded vowels ([i], [ɪ], [e], [ɛ], [æ], [æ̝], [ə], [ʌ], [ɑ], and diphthongs [ɑi], [oi]) before any other vowel. The labial constriction [ʷ] will follow rounded vowels ([y], [ø], [œ], [u], [ʊ], [o], [ɔ], [ɒ], and diphthong [ɑu]) before any other vowel. The velar constriction [ˠ] will follow nasalized vowels ([ɛ̃], [ũ], [ɔ̃], [ɑ̃]) before any other vowel.

As discussed in the orthography section, these constrictions are unmarked in the native Łedu orthography, but one will always appear where a carrier letter is present, except in word-initial conditions. The constriction does not correspond to the carrier letter itself, but instead to the features of the vowel in the preceding syllable (more specifically, the vowel diacritic on the preceding letter), since the carrier letter will always represent the second of two vowels in sequence. On the other hand, in the Latinate orthographies, the constrictions will be represented, using either <j>, <w>, or <x̂>. There is no overt way to distinguish these from the <j>, <w>, and <x̂> that represent [j], [w], and [q/ɢ] respectively; this ability comes merely from knowledge of the language. The most reliable clue if one is familiar in the least with Łedu’s various morphemes, however, is that vowel adjacency and constriction occurs strictly at morpheme boundaries, while the consonants in question can occur at morpheme boundaries, but do so rarely, tending instead to appear within atomic morphemes – and they are never conditioned because of such boundaries even when they do appear there.

/ə/-dropping

In a few predetermined morphemes, an /ə/ that is still represented orthographically (always unstressed as <ę>) is habitually dropped by the vast majority of speakers. Archaically, it was pronounced, and speakers who fancy archaisms may still realize the [ə]. The morphemes in which it is standard to drop this vowel are the postpositional endings (-gę, -dę, -dhę, -qę, -bę) provided they aren’t followed by another suffix, the verbal coordination prefixes (‘sę- and ‘vę) if consonant harmony and sonority rules allow it, the nominal coordination suffixes (-ęs and -ęv) if the same applies, the second syllables of verbal negation prefixes (vaunę-, vaushę-, vaufę-, vaukę-, etc.) if the same applies, and in the case of the indicative modal vowel in the copular suffix (-ęgıl-) if the same applies. In each such case, the onset consonant(s) of the syllable that would contain /ə/ either join the preceding coda, or become a new onset if another syllable follows that would otherwise lack an onset (or if satisfying the preceding coda forces a consonant into a following onset that already existed). In no other circumstances is it customary to drop this vowel, and in fact, doing so may cause further confusion between words.

Stress Patterns

Stress in Łedu is rather unique: primary stress falls on the first syllable of the word’s lexical root morpheme, secondary stress on the two syllables flanking it, tertiary stress on the two syllables external to those, and so forth, in a pyramid-esque pattern of ascent and descent. Lexical stress is expressed primarily by means of pitch, but also subtly in terms of length. Length reductions are never so severe, at least among most speakers, that vowels are reduced to [ə] or dropped entirely. In fact, the only allophonic change that stress effects at all is the transformation of /ə/ into [ʌ] in stressed syllables. Below are a pair of glossed examples to illustrate stress patterns, using the words rhęrręłhmübvxhözowyzęgnxıli’ (“is the one used to unlock”) and særhędhrræmöłlygrhełhekęs (“also for the counterrevolutionized”):

As for sentential stress, that matter is largely up to the speaker’s preference, and many speakers aren’t inclined to stress any sentential element, but it’s also common to stress either the verb, or an argument that is either implicitly or explicitly focused (in the latter case, using the prefix x̂ę’-). General intonation doesn’t differ significantly from most familiar cross-linguistic patterns: statements tend to be flatter, questions rise toward the end (especially yes/no- questions), exclamations are louder, parenthesized content quieter, and so forth. Łedu doesn’t place much emphasis on stress at the sentential level, but neglecting it at the lexical level is more noticeable.

Compound Words, Adjectival Endings, and Transformative Priority

Although this section presents crucial morphological information to perhaps a greater extent than it does phonological information, it also serves as a testament to how thoroughly integrated the two are. Specifically, we deal here with the role that phonological mutation plays in the morphology, and what concessions the existing phonological wordforms make in order to accommodate some of these changes.

In Łedu, many morphemes follow a predictable pattern of phonological change in order to index new meaning. For example, the timeless tense form of a verb (treated as its index form) ends universally with a voiced fricative, although said fricative’s place of articulation depends on the verb class. When that verb is employed in the present tense, however, that ending instead mutates into a voiced stop bearing the same place of articulation (except in Class 3, where it becomes a different voiced fricative). Similarly, in the past tense, the verbal ending is the devoiced counterpart to its present tense ending, and in future tense, a devoiced fricative sharing its articulatory place is appended to the past tense ending. One indexes verb tense, in other words, by manipulating the voicing and manner features of the mutable verbal ending morpheme. To go a step further, transforming this ending into a nasal stop sharing the original place of articulation nominalizes the verb (except in Class 3, where one produces a click instead), and in any case, one may then append further morphemes to signify aspect and various other properties. Verbs, however, are not the only grammatical category to which such transformations apply.

OUTDATED: The following 2 bracketed paragraphs need revision, in light of inadequacies with the morphophonological pattern described herein. [Adjectives, too, are oftentimes formed by mutating the final consonant of a nominal root (or in other circumstances, adding a new final consonant). For example, one may form the adjective of association khlûrht(oint) (renal) simply by adopting an adjectival agreement ending for the root khlûrht-, ordinarily of khlûrhtoin (kidney). However, if one transforms the final stop consonant of the root into the corresponding nasal, khlûrhn(oint), one expresses the meaning bearing kidneys as an adjective. And if one transforms said final stop into the lateral fricative of corresponding voicing, as in khlûrhł(oint), it signifies bearing the qualities of a kidney – in other words, kidney-like. One builds most adjectives in this fashion, by adapting nouns and altering their final consonants (or adding them if they lack them already), save a select core of atomic proto-adjectives, and other technologies that allow for adjectivization of verbs. This process is elaborated in rigorous and comprehensible detail in the Grammar section.

Nonetheless, in the very examples above, what determines the consonant that transforms within the root? There are two consonants in the coda of the root khlûrht- – [ɹ], followed by [t] – although in any usable form of the word, the [t] would instead join the onset of the forthcoming syllable, as in khlûrhtoin. However, we work with the isolated root when performing this process, so what allows [ɹ] staying priority, and defers the transformation to the [t] in the case of khlûrhn(oint) or khlûrhł(oint)? This is the rule that these endings transform the first stop (or lateral approximant) among the consonants at the end of the root, and then delete every consonant following it. [ɹ] is an approximant, so it is passed over, whereas [t] is a stop, so it is transformed. In the case, then, of a nominal root such as rhondærr- (willpower), which does not contain a stop in its final syllable’s ending, the appropriate nasal or lateral fricative is simply appended instead, creating an adjective such as rhondærrn(oint) (willful). And in the case of a nominal root such as xhågl- (pasture), where a consonant exists after the first stop, one wipes clean all consonants past the point of concern – for example, xhånh(ötsht) (bearing pastures). When the nominal root already ends in a nasal, as in ydläsm- (family), one replaces the final nasal with a nasal stop one degree farther back in the mouth: ydläsn(oint) (bearing family) – cycling forward again to [m] from [ŋ]. Alternatively, when transforming into a lateral fricative, if the noun root in question already ends in a lateral fricative – as in aumghrhüł- (walrus) – one switches its voicing to reflect its new adjectival status: aumghrhüłh(oint) (walrus-like). Thereby, such transformations respect a predictable pattern.]

Apropos tangentially is the business of compound words in Łedu. While so-called type compounds are created simply by appending the suffix -øx̂- to one noun and placing it before another, true compounds are welded together by way of portmanteau. For example, the type compound qzinux̂øx̂-dhochıhhtell (tooth pain) is contrasted with the true compound qzidhocholl (a toothache), which is composed from the roots qzinux̂- (tooth) and dhochıhht- (pain) mashed together. In view of this mechanism, what determines which parts of the two initial roots get combined? Generally, the determinant is quite simple: it’s just the first syllables of each. If a coda is present in the first syllable of the first root in the compound, it is included, but may mutate to harmonize with a following consonant. If the second root in the compound begins with a vowel, on the other hand, the onset of the second syllable in the first root is also included (if there is one). The second root in the compound includes the coda of its first syllable and the onset of its second, forming the end of the compound root before the nominal feature complex is appended at the end (and its first consonant may mutate to harmonize with the preceding coda if needed). One may add more sounds to the compound than just the first if leaving it to the first syllable’s onset and nucleus alone would render it indistinguishable from another compound word, but Łedu syllables are so unique that overlap is exceedingly rare. It is, however, customary to include the onset of a root’s second syllable as a makeshift coda for its first syllable in compounds, if the first syllable consists only of a single nuclear vowel and no consonants: for example, when it’s the first element in a compound, the root olushtr- (water) is reduced to ol(l)-, as in ollkėzgoin (water lizard). For the most part, true compounds transpire in this fashion. The same rules apply to adjectives, or postpositions that are compounded onto existing words like prefixes, principally onto verbs. However, certain factors can complicate the matter of compounds.

When one incorporates a preexisting compound word into a new compound, one must not shorten it to one syllable, because both syllables (or more) in its compound root express different parts of its meaning. For example, xhudzjauxlagen (government) comes from xhudzjaux- (governing) + lagęq- (body, institution), and xhudzjaux- is itself nominalized from the compound verb xhudzjauz (to govern), composed of xhuwę- (over) and dzjauz (to control). One cannot reduce the compound xhudzjaux- to *xhu- to make xhudzjauxlagen shorter as *xhulagen, because the -dzjaux- bit is meaningful in its own right! Compounded compounds are exceedingly rare, but they may occur, especially in the case of culturally-prevalent compounds in which one element has essentially been reduced to a suffix. Another example is darrlürhksånhötsh (battery cell), composed of darrlürhkrh- (electricity) + ksånh (part), the former of which is already a compound of darrva- (lightning) + lürhkrh- (energy).

Furthermore, when forming compounds from nominalized verb forms (including subnominalized verbs, when necessary), the whole nominalized form must be used because verb roots are usually irreducible: for example, krhonhėxbblürhmnoin (ox) is composed of the roots krhonhėx- (labor) and bblürhmn- (cattle), the former of which is the nominalized form of krhonhėz (to work physically). Despite being a bisyllabic root, krhonhėx- is not reduced to *krho- for purposes of a compound, nor even *krhonhė-, because it derives from a verb – instead, its whole second syllable is included, final consonant and all. Another example is nhgollmghłłæpmxõll (eulogy), derived from nhgollm- (mourning) and ghłłæpmx- (word), further from nhgollv (to mourn) – the [m] is retained rather than reducing to *ngholl-. There are a few exceptions, but these come from adjectivalized verbs: łłyfeloll (wife), for instance, comes from compounding the full adjective łłynxôrhtolld (married) onto fel- (lady), rather than the nominal root łłynxôrhn- (marriage), which would produce *łłynxôrhnfeloll (marriagelady). Only in these rare cases of circuitous adjectival etymology is contracting a verb-derived root apparently acceptable for compounding purposes.

Of final note, some prevalent nouns – known informally as culturonyms for their wide and categorical use in compounds – have been reduced essentially to suffixed roots, something like clitics, rather than true compound roots. Among these, some have incurred seemingly random phonological reductions over time, as if the number of tongues pronouncing them has worn away at their sonic matter like waves eroding rock. One example, in fact, is the very name of this language itself: Łeduwell. The compound root Łedu- is composed of the elements Łet- (Łet, the culture’s self-identifier), and durrghi(j)- (language). If durrghi(j)- were truncated normally, in the manner of most words, the name of the language would be *Łedurrghell, and we’d be calling it “Łedurrgh” throughout this site – but as a function of durrghi(j)-’s culturonymous status, it has been eroded to the sort of suffix -du- instead, lopping off the /rɣ/. Another example can be found in the names of months: Obôndaijoll, the first month of the Łetnürr year, derives its name from obôn- (comet), and daivhhäks- (month). A normal rendering of the latter word in the second compound position would produce *Obôndaivhholl, but this is not the current norm, and instead we treat it as the suffix -dai(j)-. Another property of culturonymous suffixes, one may note, is that they don’t truncate the first root in the compound unless it’s phonologically unfeasible not to: the Łet culture is the Łetnürrell, not *Łenürrell (ignoring the fact that Łet- would never contract anyway, as one does not truncate proper nouns, even in compounds), and the first month is Obôndaijoll, not *Obdaijoll. This feature further lends to comparisons with enclitic status.