An Overview of Khaling Verbal Morphology*

Abstract

This article presents a comprehensive description of the Khaling intransitive and transitive verbal paradigms, focusing on the complex system of stem alternations, which is unmatched in the Sino-Tibetan family. Using internal reconstruction, we posit a reconstructed root and a set of rules from which all attested alternations can be derived. Paradigms generated by a computerized version of these rules have been validated by several speakers. This reconstruction serves as a tool to efficiently classify and describe morphophonological alternations, and as a basis for further research on diachronic morphology involving comparison with other closely related languages such as Dumi and Koyi. [PUBLICATION ABSTRACT]

Headnote

This article presents a comprehensive description of the Khaling intransitive and transitive verbal paradigms, focusing on the complex system of stem alternations, which is unmatched in the Sino-Tibetan family. Using internal reconstruction, we posit a reconstructed root and a set of rules from which all attested alternations can be derived. Paradigms generated by a computerized version of these rules have been validated by several speakers. This reconstruction serves as a tool to efficiently classify and describe morphophonological alternations, and as a basis for further research on diachronic morphology involving comparison with other closely related languages such as Dumi and Koyi.

Key words: morphophonology, internal reconstruction, portmanteau morpheme, templatic morphology, tonal alternations, Kiranti, Khaling, Dumi

Kirat Khaling Rai Development Association4

...

(ProQuest: ... denotes non-US-ASCII text omitted.)

1. Introduction1

Despite the fact that Khaling is one the earliest Kiranti languages to have been described, its morphology is still poorly known in comparison with other neighbouring languages such as Dumi (van Driem 1993). Previous work on Khaling failed to mark tone alternations in verbs and vowel length, and gave an incomplete account of the verbal paradigms.

The present article, based on recent fieldwork, provides a comprehensive account of basic verb stem alternations in Khaling, covering the non-past, past and imperative paradigms for all categories of verbs. Reflexive, negative, and auxiliary paradigms will be treated in further publications.

This article is organized in three major parts. First, we provide a general account of the Khaling language, including sociolinguistic data, synchronic phonology, and a basic account of the verbal system. Second, we present a set of morphophonological rules, based in part on the insights in Michailovsky (1975), to derive the attested stems from reconstructed verb roots. Third, the paradigms of CVC intransitive, CVC transitive, CVCt transitive and CV roots are described in four separate sections. Two appendices provide a list of all possible stem alternations and a computer-generated list of all regular verbal paradigms, based on the rules set out in this article.

2. General overview

Khaling is a Sino-Tibetan language belonging to the Kiranti subbranch, spoken by around 15,000 people in Solukhumbu, Eastern Nepal. It is in contact with Sherpa, Thulung, Kulung and Nachiring. Nearly all speakers of Khaling are also bilingual in Nepali, but the language is still being acquired by children, even by some non-Khaling children (Sherpa, Tamang, or Kami) living in the Khaling-speaking area. The following map (Opgenort 2011:254, reproduced here with permission of the author), presents the distribution of Kiranti languages, including Khaling:

The data presented here were collected in July 2011, by a team of linguists during a month of intensive elicitation and discussion with Dhan Bahadur Rai in Paris, and in early 2012 by Guillaume Jacques during two months of fieldwork in Kathmandu and in the district of Solu Khumbu, mainly in the villages of Phuleli and Kanku.

The main existing source of data on Khaling is a dictionary by linguists of the Summer Institute of Linguistics (Toba & Toba 1975). This dictionary is a preliminary work with many errors, and thirty years after its publication its authors advised one of the authors of the present article not to rely on it. Still, it remains quite rich, with over 500 verbs, and it facilitated our search for verbs of different root categories.

On the basis of the infinitive and the 1SG>3SG form cited for transitive verbs in the dictionary, Michailovsky (1975) was able to internally reconstruct verbal roots (e.g. |lop| 'to catch') of the form (C)V(Cf)(t), with a system of five reconstructed vowels, largely confirmed by the present research, and to show that the 'high tone' (in fact a falling tone) marked on verbs in the dictionary predictably correlated with root-final (Cf) stop as indicated (usually) by the 1SG>3SG form. But the morphology remained incom- pletely described (I. Toba 1973, S. Toba 1984). The morphological paradigms presented here reveal a complex morphophonology involving stem rimes, including tones, and a previously undescribed opposition of quantity.

Khaling is a relatively homogeneous language. The data in this article is based on the dialect of Phuleli, but some marginal dialectal differences are briefly mentioned.

3. Phonology

This section presents a summary of the phonological inventory of Khaling, which is necessary to understand the verbal system.

Khaling has a rich vowel system of ten distinctive timbres. In our recent research, we found an opposition of length on eight of these.

There is no opposition of quantity on /?/ (which is always short) and /o?/ (always long) in the Phuleli dialect. In other varieties some speakers also have a long /??/ phoneme. There is no opposition of quantity in syllables with sonorant finals.

The inventory of syllable-initial consonants comprises the following 26 phonemes:

The phoneme status of the glottal stop /?/ is debatable, as words with initial glottal stop could be alternatively analyzed as having zero-initial. However, its presence is conspicuous word-internally in forms such as ?i?i 'you are angry' and word-initially in quasi-minimal pairs such as:

(1) ?o?? jn? 'to pronounce ritual words'

wo?? jn? 'to put in'

Few clusters are allowed in the syllable onset: only velar or labial stops followed by r or l are observed. No medial -j- or -w- are found. Apparent examples of medial -w- occur in realizations of the diphthong /o?/, e.g. /k?o?n? n?/ 'to go'.

Only ten consonants are found syllable-finally: p t k m n n r l s j, to which we must add the marginal phoneme /ç/. No complex codas are possible.

The segment [ç] is originally the allophone of /t/ before velars and labials. It also appears however before the 1P.PST suffixes -tiki and -t?k?. Two hypotheses are proposed to explain the appearance of [ç] in this position. First, we could propose that -tiki and -t?k? come from -ktiki and -kt?k?, forms still attested in open-vowel root paradigms, and that the -k- of the cluster was simplified in consonant-final stems. Second, these forms could be the result of analogy. If we take the verb |set| 'to kill' (cf §7), the expected 1PI>3 forms, assuming the rules *t > ç /_[+velar|+labial] would be:

(2) 1PI.NPST *set+ki > seçki

1PI.PST *set+tiki > *settiki

It is therefore possible that attested seçtiki is analogically renewed after the non-past form seçki.

Whatever its origin, /ç/ is in opposition with both /t/ and /s/ as the first element of an internal cluster before /t/ (VttV, VstV and VçtV are all possible), so it must be assigned phonemic status despite its limited distribution.

Tone is phonemic (1) on open syllables with long vowels and (2) on syllables with resonant finals. There are two tones: level and falling. We transcribe the former with a macron a? and the latter with a circumflex accent a? following the Africanist tradition.

Open syllables may be short without contrastive tone, long with level tone, or long with falling tone, as exemplified by (3) and (4):

(3) tsen? 'be spoiled (of flour)' (root |tse|)

tse?n? 'sieve' (root |tsen|)

tsê?n? 'close' (root |tsekt|)

(4) m? 'that'; 'to do (3S>3.NPST)'

m??? 'over there' (the reduced form of m?jo)

m??? 'completely (ideophone)' (used with the verb root |rumt| 'to coil')

Syllables with a resonant final have either level or falling tone, with no opposition of length. We find scores of minimal pairs of infinitives like the following:

(5) lo?? mn? 'to look for (infinitive)'

lo?? mn? 'to catch (infi nitive)'

Syllables ending in p, t, or k may be short or long. There is no opposition of tone, but we redundantly transcribe the phonetically falling tone heard on long syllables with stop finals, e.g. rê?p 'he stands'.

One further tonal phenomenon occurs in a specific morphological context. In the purposive construction, level tone monosyllabic nouns (but not verbs) have low-tone allomorphs when they function as locative-marked complements of the verb |k?ot| 'to go'. We find the following minimal set:

(6) a. b?? i ?u-g?as k?? m-bi k??s-t-?

cow 3S.POSS-grass chew-LOC go-PST-2/3S

'The cow went to chew the grass.' (from the verb k?? mn? 'to chew')

b. k?? m-bi k??s-t-?

elk-LOC go-PST-2/3S

'He went (to hunt) for the elk.' (from the noun k ?? m 'elk')

c. k?? m-bi k??s-t-?

work-LOC go-PST-2/3S

'He went for his work.' (from the noun k?? m 'work')

The tonal contrast between (a) and (c) is morphologically determined: verbs with level-toned infinitive stems have their usual high level tone before the locative suffix in the purposive construction, but lexically level-toned nouns appear with a low level tone. Such nouns may be considered to have low-toned allomorphs in this context. Minimal pairs are easily found, and have been rechecked with several speakers.4

Syllables in polysyllabic words may be stressed or unstressed. In verb forms, the verb stem is always stressed, as is the first syllable of a bi- or tri-syllabic suffix string. Given its predictable nature in verb forms, stress is not indicated in the present paper.

4. Verb morphology: the affix paradigm

Both the stem and the affixes of a Khaling verb vary throughout the tense and agreement paradigm. The stem-forms show alternations which can be related to the morphonological characteristics of an internally reconstructed root, whose derivation for each verb will be presented in §5. The affixes are drawn from a single morphological 4 paradigm which applies essentially to all verbs. The present section is a general introduction to the affix morphology.

The template in Table 3 presents a synthetic view of the affixes that may appear in a finite verb form with their relative order and position with respect to the verbal stem (Σ). The affixes are organized according to a templatic, rather than layered, principle (see Bickel & Nichols 2007).

There are one prefixal position and seven suffixal positions. Some inflectional suffixes (position +1) originate from 'trapped' morphemes in Harris & Faarlund's (2006) sense, i.e. affixes sandwiched between the stem and a clitic (which afterwards became integrated in the verbal template).

Many pairs of affixes are in relations of mutual exclusion, so that the number of possible combinations is far less than the table would suggest. In the following examples, taken from texts, the position class of each affix is marked as a superscript in the tran- scription:

(7) ts?tsts?su-?? me?s? sîn-t?-??-su? ?e

children-ERG thus ask-PST-2/3-DU HEARSAY

'The two children asked her in this way.' (Solme and Lamalit II, 3)

(8) k??m-bu-ju ts?? m?ô?-t?-?? m?? m-t?-??-si?

house-LOC-down forget-PST-1SG remember-PST-2/3-REFL

melo l?? -n1-p? 2-n? 3 g? r e ? s-t? - ? ?

at.that.time fetch-INF-REACH-INF be.reluctant-PST-2/3

'He thought that he had forgotten (his plough) in the house, but he was reluctant/hesitated to go to fetch it at that time.' (The plough, 7)

(9) g?o?? lp? g?rô?t m? n?? r-?? mu-1-ku-k??t2-w?-?? ?ola

big goat that tiger-ERG NEG-eat-COMPLETELY-IRR-2/3 maybe

'That tiger probably did not eat the big goat.' (Tiger and fern, 29)

The present article describes the non-negative indicative and imperative paradigms. Forms containing auxiliary verbs (suffixal position 2), which express either Aktionsarten or Associated Motion (go to, come to etc), will not be treated, nor will reflexive or negative forms.

Tables 4 and 5 present the non-past paradigms of the intransitive verb 'to have enough' and the transitive verb 'to catch', whose reconstructed roots (§5.1) are respec- tively |sop| and |lop|. In the transitive paradigm, the row-headers in the first column indicate the person and number of the agent and the column headers those of the patient. To simplify the presentation, the suffix string of tense and agreement markers is tran- scribed and glossed as a single unit: thus the suffix -iti is glossed as a portmanteau '1DI:PST' rather than as a combination of three morphemes -i3-t?-i?.

Note that the correlation between syntactic transitivity and conjugation class is not 100% congruent. We find one intransitive deponent verb |??t| 'to say' (1SG ??? in?) with intransitive conjugation which is syntactically transitive, as agreement occurs with an argument marked with the ergative marker -??. We also find several morphologically transitive verbs (including CVCt) which do not allow any argument in the ergative, and which we call transitive deponent. These verbs (for instance |?omt| 'to ripen', 3.NPST ?o?? mdu) only allow non-animate arguments, and are restricted to 3SG forms. In the rare cases where 3DU and 3PL forms are found, these agree with the absolutive argument. Deponent verbs will not be discussed in any detail in this article, but some examples can be found in the annex, as for some root types (such as |Comt|) only transitive deponent verbs are known.

In the paradigms, the verb stem alternant is marked off from prefixes and suffixes by hyphens. The boundary between stem and suffix does not necessarily correspond to a syllable-boundary. In particular, the final consonant of a CVC stem functions pho- nologically as the syllable-initial before a vowel-initial suffix, and as a result, the vowel of a CVC stem may occur in a phonologically open syllable.

As noted above, essentially the same affixal paradigm applies to all verbs. The transitive paradigm, in which agreement with two arguments may be shown, is in general a superset of the intransitive one. Thus inverse6 transitive forms with 1st and 2nd person object (see the row with 3SG agent in Table 5) have the same suffixes as the 1st and 2nd person intransitive forms, except that they carry the inverse prefix ?i-7 and may have an extra suffix to mark dual or plural number of the 3rd person agent. These shared or partially shared forms will be referred to as 'inverse/intransitive' below. Affixes unique to the transitive paradigm are found in 1SG>2 forms and in certain 'direct' forms with 3rd person object (e.g. 1S>3, 2S, 3S>3S) in Table 5.

In the more complete paradigms below it will be seen that all past tense forms have the marker t in the suffix string, often with a supporting vowel.

5. Verb roots and stems

After a short general introduction to stem alternations, we shall show how these can be used to reconstruct a root form for each lexical verb.

The basic stem alternations of all but open syllable roots are illustrated in Tables 4 and 5, which show the alternations in roots with the final |-p|. Stop finals in CVC roots generally show nasalization before a nasal-initial suffix (as in the infinitive), and are often voiced before a vowel initial suffix, alternations that do not affect the stems of roots with continuant finals. Among root-final stops and nasals, bilabials have the most straightforward stem derivations: dental root-finals undergo more complex changes, while velar root-finals are dropped in some forms, with compensatory lengthening.

We identify a basic alternation, affecting most rimes, between strong and weak vowels. In Tables 4 and 5 the alternation is between o? (strong) and ? (weak). The strong alternants of back root vowels are backed (here to o) before the 1S>3 suffix u. Table 6 shows the basic strong and weak vowel alternants of roots in final |-p|:

The falling tone of the infinitive and some other forms of of |lop| and |sop| (e.g. lo?? mn? 'to catch') is correlated with the stop root-final (cf. the infinitive lo?? mn? 'to look for', root |lom|), but stems with level tones or with toneless short vowels also appear in the paradigms of these verbs.

A verb may have up to ten different stem alternants. The following Table lists the stem alternants of the verb |?od| 'to bring'.

This complexity is lessened by three factors: (1) Stems have a simple structure, CV(C)(t) (with tone and length), in which only the rime, never the onset, alternates. (2) Although the set of stem-alternants of a verb may be relatively large, it can be fully determined from a much smaller number of key forms: for many transitive verbs, the 1S>3S non-past alone is enough. In other cases, one or more of the dual, the third person past and the second plural non-past are required. This implies that (3) the number of possible alternation sets is much smaller than the number of alternants in each might suggest.

5.1 Internal reconstruction of verb roots

Following the basic insights in Michailovsky (1975, in press), it is possible to posit an internally reconstructed or morphophonological root form from which the occurring stem forms can be derived.

The reconstructed root form and the transitivity type (intransitive vs. transitive) are all that is needed to assign each verb to its conjugation class. This reconstructed root has a diachronic reality - it represents, with probably some minor changes due to analogy, the pre-Khaling form of the verbs, and it is the form which should be used in comparative work. It is unlikely that this root has a synchronic reality for speakers, but the diachronic perspective greatly simplifies the classification of observed alternations. Synchronically, the reconstructed root form is the most economical statement of the information needed to distinguish one Khaling verb from another (barring homonymy), and the inventory of possible root forms (defined by the inventories of reconstructed onsets, vowels and finals) defines the space of possible lexical verb roots. In the course of our research, this allowed us to predict the existence of some conjugation types (in particular the |-ent| root type) and all their correct forms before finding actual verbs which exemplify them.

In the following sections, we shall first present the complete inventories of consonantal and vocalic alternation sets. Then, from these alternation sets (presented in §5.2 for consonants and §5.3 for vowels), we shall posit the reconstructed form of the root final consonant and vowel. Since initial consonants present no alternations, the root onset will at times be identified by a default C (representing a consonant or a C(l/r) cluster). Finally, the set of rules necessary to derive the surface stem forms from the reconstructed roots will be presented.

The stem derivation rules state the root segments or rimes affected, the result, and the morphophonological and morphological contexts that condition application of the rule. For computational simplicity, the rules are defined so as to avoid feeding or bleeding relationships. Each rule applies independently to the root if the stated conditions are met. The resulting stem reflects all of the alterations to the root contributed independently by the applicable rules.

The rules by which we propose to derive occurring stem forms from the roots do not in general belong to the synchronic phonology of the language; they are conditioned by specific morphophological contexts. They no doubt to a large extent reflect earlier phonological rules by which verbal stems were influenced by suffixes. But although we believe that our reconstructed roots give an idea of the pre-Khaling form of verbal stems, we have no reconstruction of the earlier forms of the suffixes, so the reconstruction of these phonological processes must remain a subject for future research.

5.2 Consonantal alternations

There are sixteen alternation sets of stem final consonants,8 fifteen with transitive verbs and eight with intransitives. In the following table, these are presented with a default onset C and the stem vowel e; e is chosen because it is the vowel with the fewest alternations, limited to tone changes and length.

The lack of synchronic phonological motivation for the morphophonological alternations is seen in the table. The suffixes 1SG.NPST (tr) -u and 1DE.NPST -u and (seen in the first and fourth columns in Table 8) are homophonous, but they occur with different stem alternants. Thus, one cannot describe the stems only in terms of phonetic environment: most of the alternations are determined by the morphological context, and only occur with a specific set of suffixes (The precise list of suffixes associated with each stem will be presented in the appendix).

Of all verb forms, the 1SG>3SG.NPST of transitive verbs is one of the most informative: all alternation sets (except those numbered 12 and 14 in Table 8, for which a dual form is also required) can be distinguished by this form alone.10 We reconstruct the root-final consonant (Cf) based on this form. For intransitive verbs, the 3SG.NPST is similarly useful.

In Table 9, each alternation set is associated with a reconstructed Cf (sometimes with additional postfinal t) which identifies it. The stem in 1SG>3SG.NPST (tr) differs from the reconstructed form only by voicing of the Cf, and in the case of nt-final roots by assimilation of n to the following dental.

The sixteen rules of Table 10 derive the surface forms from the reconstructed final consonant of the root:

5.3 Vowel alternations

Despite the fact that eighteen vowel phonemes occur in Khaling, only ten categories of vowel alternations are attested for consonant-final roots, exemplified in Table 11 with roots in |-k| and |-p|.

In alternation set 4, there is free variation between the realization u and ? in the infinitive and 3S.NPST (cf rule 1.10).

Each alternation set represents a single root vowel. We do not need to posit 10 different root vowels, however, because some pairs of sets (hence of potential root vowels) are in complementary distribution, as shown in Table 12.

Only alternation set 3 occurs in all contexts; the other sets fall into pairs related by complementary distribution.

Sets 1 and 2 are in complementary distribution with regard to the transitivity of the verb: set 1 is restricted to transitive and set 2 to intransitive verbs. We consider that these two sets taken together are almost in complementary distribution with set 7, conditioned by velar (sets 1/2) vs non-velar (set 7) root finals. There is one exception, however: jaln? 'to strike' (1SG j?lu) is a set 1 verb with a non-velar root-final, whose expected form would be *j?? ln? (1SG *j?lu). Compare p?? ln? 'to coil' (1D p?lu). |a| and |?| certainly both originate from one proto-phoneme *a, given their quasi-complementary distribution and the numerous alternations between /a/ and /?/ in the paradigms of intransitive |-ak| verbs and |-?| verbs. The irregular vocalism of |jal| can be explained in two ways. First, it could be due to a special dissimilatory rule of |?| after |j|. This hypothesis appears problematic however since we do find verbs with initial |j| and |?| vocalism, such as |j?t| 'to like'. Second, it could be a borrowing from Thulung jal- (Lahaussois 2003:13).

Sets 4, 5, 6 are in complementary distribution with 8, 9, and 10 respectively, with regard to the final consonant: the first group (1-6) only occurs with velar final consonants |-n| and |-k| (including the clusters |-kt| and |-nt|, cf §7), while the second group (7-10) appears with all other final consonants.

As a result, and ignoring the irregular stem rime of the verb jaln?, only five vowels need to be reconstructed in non-open verb roots (Table 13).

The following six rules of Table 14 are needed to derive the surface forms from the root:

5.4 Synthesis

The internally reconstructed verb root fits the template CVC(t). The initial C can be any one of the regular Khaling consonant initials or stop+resonant clusters (§3). The final consonants of the verb roots differ in two ways from the inventory of word or syllable codas: (1) final -s, which is attested in nouns (for instance ??s 'man'), is not attested in verb roots.12 (2) roots allow final Ct clusters.

The reconstructed vocalic inventory of the roots, with only five vowels, is con- siderably reduced compared to the 18 phonological vowels found in Khaling, but is consistent with what can be reconstructed in neighboring Kiranti languages with rich vowel systems; e.g. Bahing (Michailovsky 1975), Thulung (Allen 1975), Wambule (Opgenort 2004), and Dumi (Michailovsky in press).

The set of rules presented above suffices to generate most forms, but the three additional rules relating to vowel length and tone are also needed:

In addition, the forms of intransitive |-ak| verbs present a series of idiosyncrasies and need to be treated separately.

By combining the 24 rules presented above, one can derive stems from the recon- structed root. There are no non-alternating verbs, although some categories (such as |-er|) alternate less than others.13 Transitive CVC verbs have at most 10 stems, CVCt verbs at most 6 stems, and CVC intransitive verbs at most 5 stems.

A Perl script implementing these rules produced the list of 123 regular paradigms in Appendix B. This computational application (and the validation of the results by speakers) confirms the validity of the rules presented in this section.

Providing a genuine historical analysis of the Khaling verbal system will require a systematic comparison with the closely related Dumi and Koyi: some of the rules (for instance 1.2, 1.3) are shared with Dumi and Koyi and should be reconstructed back to proto-Dumi-Khaling, while other rules (all the vowel change rules) are probably recent innovations, that took place long after the split with Dumi. The ordering of the sound changes will require a thorough comparison, and cannot be undertaken without taking nouns into account. The analysis provided here is in fact the prerequisite for historical comparison; a comparable analysis of Dumi (Michailovsky in press) and Koyi (Lahaussois 2009) is necessary before attempting any comparison between the three languages.

6. CVC intransitive verbs

Intransitive verbs have at most five different stems. These stems are derived using the rules defined in §5. In the tables below, the rules which must be applied to derive the form for each person/number/tense combination are given. Where the rules are separated by a slash, it is because different rules apply in mutually exclusive contexts depending on the phonology of the root. The application of the rules is exemplified with two distinct root types, |-ot| and |-ur|.

To clarify the discussion of these forms, we shall give distinct labels to all the stems, which will also be used in the appendix. The labels are composed based on the rules that produced the stem.

Stems which have undergone the opening rule 2.3 (for non-velar roots) and the backing rule 2.4 (for velar roots) are called strong stems; stems that have not undergone these rules are called weak stems.

For intransitive verbs, two weak stems need to be distinguished, the basic one Σ(w) and and the lengthened (or geminated) weak stem Σ(w,l) found in the imperative singular exclusively. Dental-final root verbs have an additional past weak stem Σ(w,p) which undergoes rule 1.3.

Strong stems need to be divided into three sub-stems:

a) The plain strong stem Σ(s), which undergoes rule 1.7 in the case of dental-final roots, and rule 3.1 in the case of |-Vp| roots.

b) The assimilating strong stem Σ(s,a), which undergoes rules 1.5 and 1.6. Note that rule 1.5 applies to the 1SG, while 1.6 does not. This implies that dental roots (which undergo 1.6) have the plain strong stem in the 1SG.NPST, while other root classes have the assimilating strong stem. Velar final stems are a special case: rule 1.5 only optionally applies to the 1SG.NPST, so that for |CVk| roots both CV??n? and CV?nn? are possible, while for |CVn| roots both CV??n? and CV?nn? are possible. In the appendices, only the first variant is presented.

c) The 1P strong stem Σ(s,1P), which undergoes rule 1.4. This stem is distinct from Σ(s) only in velar and -p final roots.

The first singular and dual past tense suffixes appear with or without a leading vowel: -?t?, -iti, -utu or -t?, -ti, -tu. The latter set is restricted to |CVt| and |CVn| roots, while the former appears with all other root types.

Intransitive 2SG forms have two variants, either with a lengthened vowel or with the -je suffix. In |-t| roots, we find geminated |tsts| in imperative singular forms instead of expected *tsj.

Almost all intransitive CVC root verbs can be conjugated using the paradigms of Tables 16-18. The only exceptions are |Cak| intransitive roots, which present two conjugations.

The first conjugation is quite irregular, in that it includes forms apparently influenced by |C?| intransitive verbs. It only applies to the verb |b?ak| 'go (honorific)' and |dak| 'want', a defective verb only occuring in third person forms. The paradigm is too irregular to be generated automatically (or it would involve too many ad hoc rules):

The second conjugation is more regular, but seems restricted to the verb |dak| 'to bicker' (also |d?ak| in the southern dialect), a verb root only occurring in the collocation tsinger dâ?n? 'to bicker'. This verb presents regular forms for the dual such as 1DI d?ki, but still has irregular forms for the 2/3SG.

7. CVC transitive verbs

In the transitive paradigm, stem alternations presented in the intransitive paradigms in the previous section also occur, but a few additional stems must be added.

The transitive paradigm can be broadly divided into three sections: direct forms (with a third person patient), inverse forms (3>1, 3>2 and 2>1), and somewhat marginal 1>2 forms. The inverse forms are generally (except for |-ak| roots) identical with the corresponding intransitive forms, with the addition of the inverse prefix ?i- (thus for instance the 3S>1S and 2S>1S are identical with 1S intransitive forms except for the inverse prefix).

Certain direct forms, on the other hand, are quite distinct from intransitive ones. In this part of the paradigm, the stems are always followed by a suffix. Apart from the weak stems, strong stems and strong assimilated stems, we must distinguish five additional stems for some verbs (especially for |-ut| and |-ot| roots). The two verbs |set| 'to kill' and |lop| 'to catch' illustrate all possible stem alternations.

The non-past direct paradigm includes three stems that have no equivalent in the intransitive paradigm: the weak voiced back stem Σ(w,v,b) which undergoes rules 1.1 and 2.6 (for |-o-| and |-u-| verbs), the weak voiced lengthened stem Σ(w,v,l), which undergoes rules 1.1, 2.1 and 3.1 (in the first singular) and the weak lengthened stem Σ(w,l) which only undergoes rules 2.1 and 3.1 (in the second and third singular). In this last stem the falling tone is phonetic; as mentioned in §2, there are no tonal constrasts in obstruent-final syllables.

In the past direct paradigm, |Cvt| roots have two additional stems: the weak past back assimilated stem Σ(w,p,b,a), which undergoes 1.8 and 2.6 (in first singular), and the weak past assimilated stem Σ(w,p,a), which undergoes either 1.8, 3.1, or 3.2 depending on the final consonant.

As with intransitive verbs, the 1S>3.PST, 1DI>3.PST and 1DE>3 suffixes have two distinct allomorphs -ut?, -iti, -utu and -t?, -ti, -tu, the latter restricted to |Cvt| roots and the former appearing with all other root types.

Only |ot| and |ut| root verbs have distinct Σ(w,p,b,a) and Σ(w,p,a) stems. For example, |?ot| 'to bring' has 1S>3.PST ?ô?-t? and 3S>3.PST ??? ?-t?. Note that |aC| transitives have one particularity: rule 2.5, unlike 2.1, does not apply in the 2P>3.PST form. Thus the root |sak| 'to choose' has 1S>3.PST s?gut?, but 2P>3.PST ?isakt?nu not *?is?kt?nu as would be expected if rule 2.5 operated in the same environments as 2.1.

The realization of the forms of all ten stems can be predicted by applying the rules, but the full list of all stems for each root type is indicated in the appendix for reference.

The inverse forms, as mentioned above, are quite similar to the intransitive ones:

In the past paradigm, rule 2.5 does not apply in the 3>2S and 3>2P forms of |aC| stem verbs, so that for the root |sak| 'to choose' one finds 3>2S.PST ?isakt? instead of [dagger]?is?kt?, which would be expected if the context of rule 2.5 had been the same as for rule 2.1.

The 1S>2 forms present a special Σ(s,a) for |-k| and |-n| final roots (here |?un| 'to wait' and |phrok| 'to untie') due to rule 1.16.

In the 1S>2D/P non-past forms, |-p| final and |-m| final roots have the same forms due to the addition of the falling tone to all sonorant-final forms (rule 3.2). Thus, while the 1SG>2SG forms lo?? mn? (from |lom| 'to look for') and lo?? mn? (from |lop| 'to catch') are distinct, their 1SG>2DU and 1SG>2PL non-past forms lo?? m-su and lo?? m-nu are identical. This is the only case of homonymy in the entire paradigm between these two classes.

The imperative forms are the following:

8. CVCt transitive verbs

|CVCt| verbs have a conjugation that differs from both regular transitive and intransitives. The context of application of the rules presented above is slightly different in |CVCt| root verbs.

The weak stem only occurs in 1D/2D forms, and it is characterized by loss of the postfinal |-t| (rule 1.12), so that in these forms |CVCt| verbs are identical to the equiva- lent |CVC| transitive and intransitive forms. Thus k??sti is both the 1DI/3D.PST of the intransitive |k?ot| 'to go' and the 1DI.PST of |k?ott| 'to bring'.

Due to a different context of application of the rules 2.3 and 2.4, the strong stem appears in all forms except the dual and 2P>3.PST. However, four strong stems must be distinguished: the reduced strong stem Σ(s,r) in 1P, which undergoes the loss of the postfinal |-t| (1.13) but not most assimilatory changes; the plain strong stem Σ(s), which is the only one preserving the reconstructed postfinal |-t|; the assimilated strong stem Σ(s,a). Only in the infinitive and 2P>3.NPST; and the falling tone strong stem Σ(s,f), which only differs from Σ(s,r) by the failure to apply 1.4 and 1.9 and by the falling tone (rule 3.2).

We exemplify the |CVCt| paradigms with the two verbs |mimt| 'to remember' and |sent| 'to look at'.

The inverse, 1>2 and imperative forms are identical to those of corresponding CVC verbs; |CVnt| roots, however, have special forms due to the absence of transitive |Cvn| roots, in particular concerning the application of rule 1.9:

The 1>2 forms of |nd| verbs are the following, with a special strong assimilated falling tone stem Σ(s,a,f):

Imperatives of |nd| roots are as follows:

9. Open root verbs

Stem alternations in open roots are quite different from those in consonant-final roots. There are no weak vs. strong stems, and most of the rules stated above (except 2.1 and 2.2, both fronting rules) do not apply. Instead, a past lengthened stem is found in past 2S and 3S of intransitive verbs. All other intransitive and transitive forms present the basic stem, except for transitive |a| roots, whose paradigm is more complex.

The suffixes used with open-syllable roots are also slightly different: dual suffixes -ji and -ju, past dual exclusive -itu (instead of -utu), past 1S -n?t? (instead of -?ta), past 1P with an additional -k- (-ktiki instead of -tiki) and past 2/3P -tnu (instead of -t?nu).

Rules 2.1 and 2.2 apply to the basic stem (so that |a|, |o|, and |u| appear as ?, ? and ? respectively), but not to the lengthened stem:

The examples of intransitive |a| roots encountered so far are all impersonal and thus have a defective conjugation (mi ghr? 'the fire burns'). In forms other than the third person singular, |g?ra| is conjugated only in the serial construction |m? g?r?| 'to vomit terribly', for instance m?n?t? g?r?n?t? 'I vomited terribly'.

The imperative singular and plural have the same vowel as the 2S/3S forms:

A number of open syllable root verbs have irregularities, with idiosyncratic vowel alternations.

The irregular /e/~/a/ alternation in the paradigm of |je| 'to come down' might be due historically to the presence of initial /j/ (as in the case of the verb |jal| 'to strike'). The root |mu| 'to be' is unique in having a back vowel stem mu that appears in non-dual forms (except where the past lengthened stem appears).

Note that the imperative of 'come down' has the regular lengthened stem of |e| roots, different from the lengthened past stem.

This appears to be the only verb in the language whose imperative stem is distinct from the second person past form.

Transitive open roots have the same set of affixes except for the 1SG>3 past (-nt? instead of -n?t?) and 3>3(D). The pattern of stem alternations is slightly different, especially for |-a| roots, which have complex alternations:

The lengthened stem only appears in 2PL>3 past and 3SG>2 past forms. The 3SG>3.PST in particular has a short vowel, a difference with intransitive verbs that allows us to distinguish between these two verb categories (since the 1S.NPST has -n? for both transitive and intransitive).

For |-a| roots, no less than seven different vowels appear in the paradigm:

The imperative of transitive verbs has the lengthened stem only in the plural. The 2/3.PST stem is found in the singular form:

These are the same stem forms as those found in second person past, with a different suffix. Note that the 2>1 imperative suffix of open-syllabe transitive roots is -n?je instead of -?je with the weak stem: bi-n?je 'give it to me'. Open-syllable roots and -n final roots have the same imperative 2>1 (as well as 2/3>1S).

The inverse forms can be predicted from the above paradigms, but the 1S>2 forms present idiosyncrasies:

An additional -n- appears in all forms but the 1S>2S.

|mu| 'to do', which merges |Cu| root forms with |Ca| root forms), is the only irregular transitive verb:

For open verb roots, it is always necessary to have the 1S.NPST and the 3S.PST forms. The 1S.PST makes it possible to determine the root class and the transitivity, but in the case of intransitive verbs, this form is homophonous with the equivalent form for |CVn| roots. Thus, m?n?t? is the 1S.PST form of both |mo| 'to vomit' and |mon| 'to dream'.

10. Conclusion

The present paper provides a comprehensive description of Khaling verbal stem alternations and of the structure of the transitive and intransitive paradigms.

Explaining these paradigms however will require a systematic comparison with Dumi and Koyi. Many non-trivial alternations, such as that between -j/-ts/-s (in verbs such as 'to go' 1S.NP k?o?? j-n?, 1DI.NPST k??ts-i, 1DI.PST k??s-ti) are shared between Khaling, Dumi, Koyi, and perhaps other Kiranti languages, and should be reconstructed back to their common ancestor.

Although most alternations in Khaling seem to be relatively recent, due to the loss of final consonants or to the application of accentual rules (especially for the lengthening rule 3.1), it is conceivable some of the vowel alternations observed in the Khaling paradigms are not recent, but constitute archaisms.14

This paper describes only part of the Khaling verbal system. The negative and reflexive paradigms have not been treated, and the system of auxiliaries, whose conjugations are highly irregular, will be described in future publications.

Footnote

* We would like to thank Eka Prasad (Ramrise), Yadav Kumar (Bijame), Janaki (Majam), Rashdhan (Gaunaslaal), Dharmendra (Uramlaal), Dhan Maya (Malim) Rai for their help rechecking the paradigms. Kabimaya Rai (Majjati), Shubhamaya Rai (Majamphul), Nou Bdr. (Jetha) Rai (Ainaslaal), and Basundhara Rai for their assistance during Guillaume Jacques' fieldwork in Phuleli. This research financed by the LACITO and CRLAO research groups of the French National Center for Scientific Research.

1 The abbreviations in this paper follow the Leipzig Glossing Rules. Additionally, the following symbols are used in the description of the data and the definitions of the rules:

L: lengthening of the vowel; F: falling tone; C: any consonant; Cf: final consonant.

Internally reconstructed verb roots (see §5.1) are represented between vertical bars | |.

2 In the Leipzig Glossing System, this notation indicates a transitive form with fi rst singular agent and third singular patient.

3 The 1SG form cited for intransitive verbs provided no more information than the infinitive.

4 For instance s?? n -bi kh? st? 'He went to ask' from the verb root |sin | 'to ask' vs. s ?? n -bi kh? st? 'He went for the firewood' from the noun s?? n 'tree, firewood'.

5 The reflexive suffix sometimes appears between the suffixal slots +2 and +3 as in m?? m-si?-n?3-t?-?? 'I thought'.

6 'Inverse' refers to forms of the transitive paradigm in which the patient outranks the agent on the person hierarchy 1>2>3; the converse is 'direct'. 3>3 forms in Khaling are direct forms. True direct/inverse systems are found in related languages, such as Rgyalrong (Sun & Shi 2002, Jacques 2010).

7 This prefix has two overlapping functions: it marks all forms with 2nd person arguments except 1>2 forms, and it marks all inverse forms. For a historical account of the origin of this prefix, see Jacques (2012) and DeLancey (2011).

8 Open root verbs, with no final consonant, will be treated separately, in §7.

9 Verbs with CVCt roots (lines 9-16) are all transitive and thus do not have the intransitive form of column 2. Alternation set No 6 is not attested with transitive verbs, hence the grey cells in columns 1 and 7.

10 1SG>3SG.NPST is the form listed (with the infinitive) in Toba & Toba (1975).

11 As mentioned in footnote 1, F and L refer to changes to the root vowel seen in the stem (F=falling tone, L=lengthening) in addition to changes to the final consonant. Note that changes in the quality of the vowels in the stems are discussed separately in the next section.

12 Ancient -s root verbs have become open-syllable roots. For instance, |k?e| 'to steal', has an irregular derived noun k?esp? 'thief', which preserves a trace of the final consonant.

13 The only changes to the stems for |er| are tone changes and lengthening.

14 In particular, the puzzling i/u? alternation of the open-stem paradigms reminds one of the unexplained o/i alternation found in Situ Rgyalrong in the verb 'to come' pi, po (Lin 2003: 254). It is significant that this alternation occurs in the Khaling verb /pi/ 'to come (horizontal plane)', the probable cognate of Situ pi.

References

References

Allen, Nicholas J. 1975. Sketch of Thulung Grammar. Ithaca: Cornell University.

Bickel, Balthasar, and Johanna Nichols. 2007. Inflectional morphology. Language Typology and Syntactic Description, Vol. 3: Grammatical Categories and the Lexicon (2nd edition), ed. by Timothy Shopen, 169-240. Cambridge & New York: Cambridge University Press.

DeLancey, Scott. 2011. Notes on verb agreement prefixes in Tibeto-Burman. Himalayan Linguistics Journal 10.1:1-29.

Driem, George van. 1993. A Grammar of Dumi. Berlin & New York: Mouton de Gruyter.

Harris, Alice C., and Jan T. Faarlund. 2006. Trapped morphology. Journal of Linguistics 42.2:289-315.

Jacques, Guillaume. 2010. The inverse in Japhug Rgyalrong. Language and Linguistics 11.1:127-157.

Jacques, Guillaume. 2012. Agreement morphology: the case of Rgyalrongic and Kiranti. Language and Linguistics 13.1:83-116.

Lahaussois, Aimée. 2003. Thulung Rai. Himalayan Linguistics Archive 1:1-25.

Lahaussois, Aimée. 2009. Koyi Rai: a initial grammatical sketch. Himalayan Linguistics Archive 4:1-33.

Lin, You-Jing. 2003. Tense and aspect morphology in the Zhuokeji rGyalrong verb. Cahiers de linguistique - Asie orientale 32.2:245-286.

Michailovsky, Boyd. 1975. Notes on the Kiranti verb (East Nepal). Linguistics of the Tibeto-Burman Area 2.2:183-218.

Michailovsky, Boyd. (in press). Internal reconstruction of the Dumi verb: lexical bases and stem formation. Linguistics of the Tibeto-Burman Area 35.2.

Opgenort, Jean Robert. 2004. A Grammar of Wambule. Leiden & Boston: Brill.

Opgenort, Jean Robert. 2011. A note on Tilung and its position within Kiranti. Himalayan Linguistics Journal 10.1:253-271.

Sun, Jackson T.-S., and Danluo Shi. 2002. Caodeng Jiarongyu yu 'rentong dengdi' xiangguan de yufa xianxiang [Empathy hierarchy in Caodeng rGyalrong grammar]. Language and Linguistics 3.1:79-99.

Toba, Ingrid. 1973. The Khaling verb. Nepal Studies in Linguistics 1:1-14.

Toba, Sueyoshi. 1984. Khaling (2nd edition). Tokyo: ILCAA, Tokyo University of Foreign Studies.

Toba, Sueyoshi, and Ingrid Toba. 1975. A Khaling-English, English-Khaling Glossary. Kathmandu: Summer Institute of Linguistics and Institute of Nepal and Asian Studies.

[Received 14 December 2011; revised 6 June 2012; accepted 20 June 2012]

Appendix

(ProQuest: Appendix omitted.)