Two plant-based numeral classifiers in Nuristani languages: grain and branch

3.1.1 Askun

The word pol ‘grain’ was first recorded by Morgenstierne (1929: 271); and in a later glossary he listed it in the form pʋl ‘grain’ (Morgenstierne 1934: 102). Although the Ashkun language is the least well-described Nuristani language, there is evidence that this word is used as a count word for small, round, grain-like objects. Strand (2011c) lists an extended form, pol’ak ‘granule’ from the more conservative, saňuviri dialect, adding that this lexeme also has a function as “measure for grain and grain-like objects”. Similar extended forms (with suffix variants -Vg) also occur in Waigali and Prasun.

3.1.2 Kati

Next to Ashkun pol ‘grain’, Morgenstierne (1929: 271) also mentioned the corresponding Kati lexeme, pul ‘corn, grain’, noting that it may be a loanword “on account of Kati l”, however, later he also mentioned the Kati form pyuŕ (Morgenstierne 1934: 102). The form pul occurs in Grjunberg’s (1980: 284, §312) description of Western Kati numerals, but Grjunberg did not mention that pul could be a count word or classifier. My analysis in (7) makes it transparent that pul occurs in a lexico-syntactic role and functions as a numeral classifier for the word meaning ‘bullet’.

In both kamviri and katavari, Strand (2011a, 2011b) recorded pʹul with the meaning ‘granule’ and added that it is also used as a “measure for grain and grain-like objects”. Furthermore, pʹul also appears under a separate lemma (in both dialects) with the meaning ‘pimple’, and Strand connects these lemmata (‘granule’ and ‘pimple’) with OIA phála ‘fruit; grain’ (Turner № 9051). So, the two nouns are not homonyms, but they represent a case of polysemy, which must have emerged by semantic extension from ‘grain/granule’ to ‘pimple’. This semantic development can be observed in Waigali and Prasun as well.

3.1.3 Waigali

The lexeme pol appears in Morgenstierne’s Waigali glossary (1954: 286) with multifarious meanings: ‘single grain, egg, piece; bullet’, which obviously points to the circumstance that the designated objects share the conceptual feature ‘small, round items’. Although Morgenstierne did not mention that this word has a lexico-syntactic function, some of the expressions listed in his glossary reveal that pol can function as a numeral classifier, as shown in (8a–8c).

Semantic extension has also operated in Waigali, so the lexeme pol in the Nishey dialect – just like in Kati – is attested with meanings such as ‘pimple’ and ‘testicle(s)’. Degener (1998: 510) gives the following meanings for pol in the Nishey dialect: 1) ‘grain; testicles’; 2) count word (general); 3) ‘point (score in game)’. The numeral classifier function is demonstrated by expressions such as e pol gum ‘a grain of wheat’, e pol kāċ ‘a grain of millet’. The semantic scope of pol as classifier is also extended, as is shown by the following examples: ṣṭaš pol ċo ‘18 holes’, e pol üst’üm ‘one pillar’; and it can be applied even to animates: e pol zag’a ‘one child’.^[12] Strand (2011d) confirms the classifier function, stating that Nishey pol is a “measure for grain and grain-like objects”. Next to pol, Degener’s glossary gives the lemma pol’ag ‘small piece’, which is obviously a suffixed variant of pol (compare the Ashkun pair pol vs. pol’ak). The extended form, pol’ag is also attested in a classifier function: dü pol’ag pay’sa ‘two coins’.

The classifier pol has anaphoric use as well, and for this we have textual evidence from the description of a stone-throwing game, as shown in (9):

When further elements are added to a numeral classifier phrase, the numeral–classifier word order is preserved:

The classifier pol shows two seemingly peculiar occurrences in a classifier slot where a different classifier would be expected. The first one is in the expression e pol gröṣ (one–cl _grain –billygoat) ‘one billygoat’ (11).

Degener (1998: 225) labelled the classifier pol in this phrase as “Temporalklassifikator”, i.e. a classifier, which features only temporarily next to a noun that would otherwise take a different classifier (in this case, cl _head would be expected before gröṣ ‘goat’). Textual evidence, however, reveals that the phrase e pol gröṣ is in fact part of a syntagma in a narrative about a skinned goat, so pol in this case is the general classifier and refers the goatskin, not to the animal itself, as shown in (11).

The second “unexpected” case of the classifier pol occurs with the noun ḍugura ‘small house, hut’,^[13] see (12a), despite the fact that in Waigali (as well as Kati),^[14] houses, rooms or built structures are counted with the classifier pur (Degener 1998: 224) as shown in (12b) and (12c).

It seems difficult to explain why pol ‘clf _grain ’ is used in (12a). Judging by the context, perhaps pol serves the pragmatic purpose of intensifying smallness, since the narrative relates the marriage story of a poor boy, who cannot offer his beloved girl a house to live in but then builds a small hut for them. The use of pol ‘clf _grain ’ probably emphasizes the idea that what the young man built was not a proper house that could be considered (and classified) as a building, but just a modest hut, a shed. Using a classifier different from the grammatically expected one (e.g., clf.general instead of clf.human) is well-known in numeral classifier systems of East Asia, as well as Nepali, where classifier switch can be applied “to convey affective messages” (Allassonnière-Tang and Kilarski 2020: 144).

3.1.4 Prasun

A word pola ‘seed, grain’ was recorded by Buddruss from an inhabitant of the village Dewa (Buddruss and Degener 2015: 742). In other Prasun speaking villages (Zumu, Pronj, Iṣṭewi, Paṣki) the extended form dominates, and it occurs in free variants such as pul’og, pul’ug, pol’og with two meanings: ‘grain’ and ‘part, group, team’ (Buddruss and Degener 2015: 743). The form variants do not appear to be tied to one of the meanings. Although “classifiers are used sporadically” in Prasun, pul’og appears in the function of a numeral classifier for small objects or members of a group (Buddruss and Degener 2017: 28). The example in (13) shows that word order in Prasun corresponds to the general pattern in Nuristani languages: the numeral precedes the classifier.

However, a short list of examples with pul’og (and form variants) in numeral phrases from the Paṣki dialect (extracted from Buddruss and Degener 2015: 743) shows an apparent idiosyncrasy of Prasun word order: the word pul’og is not adjacent to the numeral but follows the noun:

Word order in these examples seems typologically unusual: the numeral and the classifier should always be contiguous since they have a morphosyntactic tie, so the noun designating the classified item(s) is not supposed to come between them.^[15] Moreover, how is it possible that Prasun has both types of syntactic arrangement: Num–pul’ug–Noun versus Num–Noun–pul’ug? This becomes possible if pul’ug, separated from the numeral and postposed to the noun, is not a real classifier but a class noun (cln). This is confirmed by the fact that the postposition -wag can be suffixed to it, see (14).

So, in the expression lü pul’og mülüg’eir (Num-clf-N) ‘two gemstones’ pul’og has a syntactic tie with the numeral, and thus it functions as a classifier. But separated from the numeral, as in the phrase atig kuml’ig pul’ig ‘one mulberry’, pul’ig is a class noun (cln) in close apposition with the noun ‘mulberry’. This dual treatment of Prasun pul’ug as classifier and class noun is not at all exceptional. DeLancey (1986: 439) explained that “the syntactic category of classifier is quite fluid”, and this can complicate the task of identifying classifiers. He described this fluidity by setting up a continuum ranging from pure classifiers (which are used exclusively as classifiers) to pure nouns (which never play a classifier role). Between these two extremes we find autoclassifiers, class terms (which I would rather call class nouns, and which generally occur in compounds as the head with some classifying capacity) and “nouns, which can feature in more than one function, either as a classifier or a class term, depending on their syntactic position” (DeLancey 1986: 439). The fluid nature of classifiers can help us resolve the situation observed in the behaviour of Prasun pul’og.

3.1.5 Etymology

Morgenstierne (1929: 271) compared Ashkun pol ‘grain’ with Waigali pōl ‘fruit’, as well as Kati pul ‘corn, grain’, but – “on account of Kati l?” – he judged Kati pul to be a loanword. In the same entry, Morgenstierne mentioned parallels from Dardic languages: Khowar polok ^[17] and Pashai (Degano) på, to which we can add Shumasti phäl ‘a single grain’ (Turner № 9051), Kohistani phʌ́l ₂ (Zoller 2005: 289), Kalasha pháḷak, (Birir) pháḷúk ‘grain, cereal’ (Trail and Cooper 1999: 238), pha’ɫik ‘grain’ (Heegård Petersen 2015: 265).^[18] More importantly, there is some evidence for ‘grain’ used as a count word in Dardic:

1. Pashai wāl (< *phal) is a numerative for counting grain; (lauṛowānī dialect) sātə wāl gum ‘7 grains of wheat’ (Turner № 9051);

2. Kalasha phaḷík ‘a single grain, seed, kernel’ is used for counting grains, berries, cowrie shells and beads (Trail and Cooper 1999: 238);

3. Palula phalúuṛu ¹ ‘grain’ cognate with phalúuṛu ² (adj. masc.) ‘sole, only’ as in the expression a phalúuṛi dhií (one sole daughter) ‘only one daughter’ (Liljegren 2019).

Furthermore, the Iranian language, Parachi (spoken in the vicinity of Nuristan) has phōṛ ‘fruit, grain’ as an itemizer (Kieffer 2009: 696), as opposed to the more frequent Iranian classifier dane ‘grain’. This may be a Nuristani loan in Parachi.

Degener (1998: 510) and Strand (2011d) point out the lexical connection with Indo-Aryan referring to OIA phála ‘fruit’, RV., *‘grain’ (in suphálā ‘well-fruited’, said of furrow RV.), ʻseed of a fruitʼ Amar. (Turner № 9051). Strand’s (2013) etymological lexicon provides the Proto-Nuristani reconstructed form *pal based on Kati, Ashkun and Waigali data. Prasun pul’og ‘small objects, members of a group’ is obviously a loanword because Proto-Nuristani *pV- in Prasun would be lenited to a semivowel w. Since Prasun is known to have borrowed numerous words from Kati, this might be just another Kati loanword. For the same reason, the Prasun (Dewa) word pola ‘seed, grain’ is also considered a loanword from Kati (Buddruss and Degener 2015: 742). However, in Prasun – unlike in Kati – the extended form dominates, and it occurs in free variants such as (Zumu, Pronj, Iṣṭewi, Paṣki) pul’og, pul’ug, pol’og. These forms resemble rather the extended forms in other Nuristani languages: Waigali pol’ag and Ashkun pol’ak. It is peculiar, though it may not be accidental, that all examples of classifier use in the Prasun language come from the dialect of Paṣki. This is the southernmost village in the north–south string of villages in the Parun valley, so Paṣki may be the most exposed to the influence from the other Nuristani languages, primarily western Kati.

3.1.6 Summary: grain in Nuristani

I summarize the available evidence for the polysemous lexeme grain in Table 1, which also shows (in boldface) in which Nuristani languages this lexeme has obtained a morphosyntactic function in counting small, round objects, especially items that are conceived of as grain-like objects.

Three Nuristani languages, Ashkun, Kati and Waigali have a cognate lexeme meaning ‘grain’, with semantic extensions to small round objects such as ‘pimple’, ‘testicle’, ‘egg’, ‘bullet’. The Prasun word for ‘grain’ is a loanword, and its morphosyntactic function as a classifier for small objects must have been adopted part and parcel with the lexical borrowing. The Prasun lexeme may have been borrowed from Kati (as would be the more typical case) or rather from Waigali (as appears to be phonetically more probable). I agree with Strand (2013) reconstructing a Proto-Nuristani *pal(-) corresponding to OIA phála- ‘fruit’, RV., *‘grain’ (Turner № 9051).

Since the classifier function of this lexeme is not attested in Indo-Aryan (except for a few Dardic languages), I would propose that its grammaticalization into a classifier is an innovation native to Nuristani. This morphosyntactic feature probably spread from Nuristani to Dardic. The occurrence of grain as a classifier is known from the neighbouring Iranian languages of the Pamir, but there a different lexeme appears in this role, e.g., Shugni důnā, Sarikoli duno, Wakhi δəng, which are labelled as numeratives by Rastorgueva and Èdel’man (2003: 449), noting that the Shugni and Sarikoli lexemes are loanwords. grain as a classifier in Nuristani is most likely to be an innovation motivated by code-copying, although the exact source (grammatical model) cannot (yet) be identified.

3.2.1 Semantic extensions of ‘branch’

Based on the semantic distribution observable in Table 2, ‘branch’ appears to be the primary sense since it is attested in all the languages. From this primary sense secondary meanings such as ‘teat’, ‘irrigation channel’ have emerged, and ‘branch’ was semantically extended even further to the abstract sphere of human relations (‘descendant’). Waigali, Prasun and Kati share the same pattern of meaning extension from ‘branch’ to ‘teat’, so the polysemy must be archaic in Nuristani. The polysemy and the relatively wide semantic scope of this lexeme can be interpreted as a result of metonymic extensions, since “metonymy is a regular source of polysemy” (Langacker 2013: 70). The salient visual image of tree branches is that they split into smaller branches, or – from the reverse vantage point – smaller branches merge into bigger ones creating a seam where they join. This is the foundation for the image schema that the arrangement of body parts resembles the shape of a tree. The partonymic congruence provides the opportunity for metonymic transfer, which can work in both directions: it can start from human body-part names and lead to objects (e.g., leg of a table) and abstract things (e.g. leg of a journey), or in reverse, from botanical terms to the human body.^[19] This latter direction characterizes Nuristani. Based on the pattern of compounds with tree names (such as Prasun umʹū ċā ‘branch of a walnut tree’), the lexeme meaning ‘branch’ combines with body-related nouns to form names of body parts, e.g. Prasun buṣ ċā (lit. chin+branch) ‘(lower) jaw’ (Buddruss and Degener 2015: 633) (P.) wustū ċāw, (P.K.D.I.) wustū ċā (lit. breast+branch) ‘rib’ (Buddruss and Degener 2015: 861). In the context of animals the transfer of ‘branch’ refers to udder and teat, while in the inanimate sphere the reference is to the branching of a watercourse.

In terms of kinship ‘branch’ is attested in the Waigali (Nishey) compound abr’āy-ċāw (lit. brotherless+branch),^[20] which refers to the portion of property that a brotherless woman receives from her family (Degener 1998: 366). This is comparable with the Kati (katavari) compound mʹâlćov meaning ‘mother’s agnates’ (Strand 2011b). In the kamviri dialect of Kati, ćʹoa refers to an agnatic group, e.g. in the expression ćʹoa špâ,^[21] which means to ‘cause an agnatic descent group to split by violating the rule of exogamy’ (Strand 2011a).

3.2.2 Reconstructing the Proto-Nuristani etymon for ‘branch’

Despite variant representations in transcription as ċ/ć/c used by the different sources,^[22] the initial consonant can safely be identified – as well as reconstructed for Proto-Nuristani – as the voiceless dental affricate [ʦ]. The initial s in Ashkun sau is a regular development from the Proto-Nuristani dental affricate. The symbol † before Prasun (Paṣki) ċō indicates that this is a word form which was recorded by Georg Buddruss in 1956, but during his later field work of 1970 speakers did not confirm knowing this word in this from (see Buddruss and Degener 2015: 640). I would explain this discrepancy by the circumstance that Paṣki is the southernmost village in the Parun valley and its inhabitants are the closest to the southern neighbours, who are speakers of (the western katavari dialect of) the Kati language. Thus, exposed to contact with the western dialect of Kati, the Prasun speakers in Paṣki may have used (and later abandoned) ċō as a temporary variant influenced by Kati ċov. Prasun also poses the question of how to interpret the [w] ∼ [g] variation in the forms ċāw/ċaw’ә and ċāg/ċʹāga.

I consider the disyllabic Prasun forms (ċaw’ә ∼ ċʹāga) to be the more conservative forms. The monosyllabic forms of the other Nuristani languages must have emerged via apocope. In Kati the monosyllabic form *ċ’aw (< ċaw’ә) has its newly accented [a] become [o]: ċow; this rounding process is regular both in kamviri and katavari (see Strand 2010, 5.b.iv.). In Ashkun the semivowel/w/in final position was vocalized, so *ċaw > Ashkun cau (> sau). The same vocalization is attested in the kamviri dialect of Kati, which has the variants ćʹov and ćʹoa.

Some of the Nuristani word forms listed in Table 2 were already included in Turner’s (1962–1966) dictionary of Indo-Aryan languages under entry № 12376, śā́khā ‘branch’, and in the Nuristani lemmata ćʹoa/ćʹov ‘branch; seam’. Strand also refers to Turner’s OIA śā́khā and gives a Proto-Aryan etymon *čâkhai- ‘branch’. However, Strand finds the comparison with śā́khā questionable because the sound correspondence between (Indo-)Aryan [-kh-] and Nuristani [-v] is perplexing. A further dilemma is posed by Prasun, which displays a phonetic variation [-w-] ∼ [-g-]: (P) ċaw’ә ‘branch, descendant’ ↔ (P.I.) ċ’āga ‘branch, twig, descendant’ (Buddruss and Degener 2015: 640, 641).

It seems to me possible to resolve the phonetic issue of Nuristani [-v] and Indo-Aryan [-kh-] as well as the Prasun doublets if we look at the peculiar inflectional development of this lexeme in Indo-Iranian and Armenian. Both these Indo-European branches show that the direct and oblique case stems could evolve differently due to phonotactic reasons. The evolution of nominal inflection from Old to Middle Persian produced “doublettes” that “continue two different protoforms” (Cantera 2009: 18): one goes back to the direct case (which had a monosyllabic ending) and the other to the oblique (with a bisyllabic ending). In Armenian there is a lexical dialect doublet pair, which emerged in this manner: cՙax ‘branch; firewood; brushwood’ continues the direct stem but cՙakՙ ‘a thorny plant; harrow’ reflects the oblique stem. According to Martirosyan (2009: 620) both Armenian -x in cՙax and Skt. -kh- in śā́khā- emerged by the generalization of the oblique stem. Table 3 provides a comparison of the Armenian and Iranian developments.

The phonetic split of the stem consonant in the paradigm can have two subsequent developments: the alternation was either preserved, sometimes creating lexical doublets (as in Armenian dialects), or analogically levelled out (as in Sanskrit). Both tracks of development seem to be reflected in Nuristani languages. Prasun – just like Armenian – shows a lexical doublet: (P) ċaw’ә ‘branch, descendant’ next to (P) ċ’āga ‘branch, twig’ and (I.) ċ’āg ‘descendant’. My hypothesis is that the Prasun form ċ’āga continues PNur. *ċ’ākā (with intervocalic lenition of *k), while Prasun ċaw’ә generalized the PNur. oblique stem *ċaɣ’a- with further lenition of the intervocalic fricative to approximant *ċaɰ’a-. The PNur. oblique was generalized in the other Nuristani languages: Kati ċow, Waigali ċāw, Ashkun ċau. Table 4 gives an overview of the developments in Nuristani.

To sum up, I posit a PNur. *ċ’ākā ‘branch’, which yields a lexical doublet pair in Prasun. The mechanism of doublet formation is due to the generalization of the oblique stem with a stem consonant different from that of the nominative stem. This paradigm split is also attested in Iranian (Middle Persian) and Armenian; moreover, it has produced doublets in Armenian dialects. Kati, Waigali and Ashkun preserved only the oblique stem (PNur. *ċāx’ā-) with -w resulting from gradual lenition of *x > *ɣ > *ɰ > w. This explanation reconciles the apparent phonetic contradiction between Nuristani forms with w and Skt. -kh- in śā́khā-.

3.2.3 The grammaticalization of branch into class term and classifier

As a recurring component in compounds, ‘branch’ is bound to become semantically bleached and functionally loaded, which opens the path to be grammaticalized as a class term. Since “class terms are a major source for the development of new classifiers” (DeLancey 1986: 440), the grammaticalization process of ‘branch’ in Nuristani must have reached its climax with syntacticization. This means that the new, morphosyntactic role is overtly signalled by word order change: ċā precedes the classified noun because its conceptual link becomes stronger with the numeral rather than with the noun. As a class term, ċā prominently combines with names of body parts, e.g. Prasun buṣ ċā (lit. chin+branch) ‘(lower) jaw’, and this is reflected in its classifier function attested primarily in phrases quantifying body parts, e.g. in Prasun:

The use of the Prasun classifier ċā is then extended from body-part names to the names of clothing pertaining to the respective body parts. This can be observed in Prasun, where the classifier ċā occurs with the items of clothing, e.g. (P.) čpū ċā wәzәl (= four clf _branch shoe) ‘four (pairs of) shoes’. Here it may be interesting to note that when half a pair of shoes is meant, in Waigali (Kegal) the classifier grain is used: e pol wacā (= one clf _grain shoe) ‘one shoe (= half a pair)’. Due to the limited amount of available data, it cannot be established if the difference in classifier use is motivated by the conceptual distinction between pair and single item. The classifier branch in Nuristani languages is also attested referring to length of cloth, e.g. in Kati (kamviri) ćʹoa, ćʹov ‘measure for blankets’, in Prasun (I.) atig ċā čū ^[24] ‘one end (of a turban).

In Nuristani we can also trace the conceptual parallel of the branching schema, which then is transferred to the concept of branching out/protruding and an abstract sense of splitting. Thus, in Prasun ċā ‘branch’ is used as a classifier for extremities (limbs branching from the torso) and then for items of clothing associated with these body parts. The dominant association with paired items allows for the extension to ‘eyes’. Another track of extension is to names of body parts that seem to branch out/protrude (such as jaw, rib, breast, teats).^[25] The image of splitting into branches is then transferred to nouns that are not body-related (e.g. irrigation channels) and even to expressing an abstract division like ‘branch of a family’. These conceptual transfers leading to the semantic extension of the scope of this Prasun classifier can be captured in the radial category model shown in Figure 2.

Figure 2:

The radial category model of semantic extensions of Nuristani branch.

Although there is a common lexeme for ‘branch’ in Nuristani languages, we have evidence for its classifier use only from Prasun and Kati, and no such function has been reported from the southern Nuristani area (Waigali and Ashkun). Table 5 summaries the data.

3.2.4 A note on the collective meaning of the Prasun classifier ċā

The Prasun classifier ċā can combine with another lexeme that itself functions as a classifier: (P) dam ^[26] (Buddruss and Degener 2015: 651) or its variants (P) dram (Buddruss and Degener 2015: 652) and (Z., K., Pr., I.) ḍam (Buddruss and Degener 2015: 655–656), as in the expression (P) atig ċā dam ‘one branch’. Since initial ḍ is a regular reflex of *dr- (parallel to ṭ < *tr-), Prasun ḍam is obviously cognate with Kati (kamviri) drüm, (katavari) drum ‘broom made of twigs’ (Strand 2011a, 2011b gives the meaning as ‘plant used for brooms’) and Waigali (Nishey) drü ‘broom’ (Degener 1998: 421; Strand 2011d), and ultimately must be related to OIA drumá ʻtreeʼ (Turner № 6637). When in Prasun ċā is combined with dram, the phrase conveys a collective meaning: (P.) ċā dram, (I.) ċā ḍam ‘branches’. In such constructions dram can be understood as a semantic repeater, which serves the function of collectivizing. The collectivizing sense is also attested in the context of body-part names, e.g. (I.) ċā ḍam means ‘branches (coll)’ as well as ‘extremities and head’; (K., D.) ċā ḍam ‘arms and legs’, ċā ḍam wulūm’aso әnd’eš ‘my limbs hurt’.