Nepali grammatical tradition accepts the deletion of the suffix <nu> from the citation form of a verb to derive roots in Nepali, e.g:
1a. Citation forms: khanu ‘to eat’, gʌrnu ‘to do’, aunu ‘to come’, dinu ‘to give’, dhunu ‘to wash’, birsinu ‘to forget’, sʌhʌnu ‘to endure’ and duhunu ‘to milk’
1b. Root forms: kha ‘eat’, gʌr ‘do’, au ‘come’, di ‘give’, dhu ‘wash’, birsi ‘forget’, sʌhʌ ‘endure’ and duhu ‘milk’
Nepali phonology does not supply nasalized <o> (and there are a few cases of change of nasalized <e> to <i> in the data) on the surface. This constraint has two implications: the citation suffix <nu> may have its underlying form <*no> and the roots ending in mid vowels like <*de> and <*dho> may have assumed the surface forms of <di> and <dhu> respectively before <nu> due to the prenasalized spread of [nasal] feature. The following forms of the mid vowels in the nonnasal environment may hint to the presumed underlying forms:
2a. hunu ‘to be’: ho ‘is’ (equative), hola ‘maybe’, hos ‘Let it be!’, hou ‘you are’,
2b. chunu ‘to touch’: cho ‘Touch!’, choyo ‘He touched’, chola ‘He may touch’
2c. dhunu ‘to wash’: dho ‘ Wash!’, dhoyo ‘He washed’, dhola ‘He may wash’
2d. dinu ‘to give’: de ‘Give!’, linu ‘to take’: le ‘Take!’ (in some dialects)
These data are against the traditional <nu> Deletion Rule to derive Nepali verb roots.
<hunchʌ> ‘becomes’, <hune> ‘becoming’, <hu)dʌi> ‘being’
In the majority of languages imperative singular form coincides with the root form. This strategy correctly supplies mid vowels in the roots in (2):
3a. <ho> ‘Be!’ 3b. <cho> ‘Touch!’ 3c. <dho> ‘Wash!’ 3d. <de> ‘Give’ 3e. <le> ‘Take!’
This strategy also helps to derive most of the root forms, eg:
4. <kha> ‘Eat!’, <gʌr> ‘Do!’, <au> ‘Come!’, <de> ‘Give!’, <dho> ‘Wash!’, <birsi> ‘Forget!’, <sʌhʌ> ‘Endure’ and <duhu> ‘Milk!’ (Dialectal variant)
However, it does not derive diphthong final roots like:
5a. <aunu> ‘to come’ : <a> ‘Come!’, <gaunu> ‘to sing’: <ga> ‘Sing!’, <chaunu> ‘to roof’: <cha> ‘Mend the roof!’, <daunu> ‘to train bullocks to plough’: <da> ‘Train the bullock!’, <dhaunu>: ‘to attend to’: <dha> ‘Attend to!’, <paunu> ‘ to get’: <pa> ‘Get it’
5b. <piunu> ‘to drink’ : <pi> ‘Drink!’, <siunu> ‘to sew’: <si> ‘Sew!’
5c. <lyaunu> ‘to bring’: <lya> ‘Bring!’, <nuhaunu> ‘to bathe’: <nuha> ‘Have a bath!’
Typically, this group of roots undergoes monophthongization by deleting the second member of the diphthong. This situation labels the imperative singular strategy not a general rule of root derivation in Nepali.
Deletion of <la>, the probabilitative morpheme, from a 3rd person singular form is another possibility of root derivation of verbs in Nepali, eg:
6a. <khala> ‘may eat’, <gÃrla> ‘may do’, <dela> ‘may give’, <dhola> ‘may wash’, <aula> ‘may come’, <siula> ‘may sew’, <birsela> ‘may forget’, <sÃhÃla> ‘to endure’ and <duhula> ‘to milk’ (Dialectal form)
This strategy is also found problematic in a few roots. The following table shows that different strategies may give different outputs in a few roots:
6. Rules C-ending h-ending CC-ending V-ending Vv-ending
<nu> Deletion: gʌr- sʌhʌ-/duhu- birsi-/birsa- di-/hu-/dhu- au-/siu-
Imperative sg: gʌr- sʌhi-/duhi- birsi- de/ho/dho a-/si-
<la> Deletion: gʌr- sʌhʌ-/duho- birse- de-/ho-/dho- au-/siu-
7a. duhu-/duho-/doho-/ duha-/ duhi-/duhe-
b. sʌhʌ-, sʌhi-
8. birsi-, birsʌ-, birse-
9. de-, di-
None of the three strategies mentioned above can resolve the problem of root allomorphy of the vowel ending roots.
To resolve this problem of allomorphy we can propose a fourth strategy, the mathematical strategy of HCF (Highest Common Factor). According to this strategy if all the verb forms of a root are taken together and calculated the HCF, it will generate the root form and this will be the general formula of verb root derivation in Nepali. Thus root forms of the verbs listed in (1b) will be as follows:
10a. kha ‘to eat’, gʌr ‘to do’, au ‘to come’, de ‘to give’, dho ‘to wash’, birs ‘to forget’, sʌh ‘to endure’ and duh ‘to milk’,
10b. Citation forms: khanu ‘to eat’, gʌrnu ‘to do’, aunu ‘to come’, dinu ‘to give’, dhunu ‘to wash’, birsinu ‘to forget’, sʌhʌnu ‘to endure’ and duhunu ‘to milk’
10c. hunu ‘to be’: ho ‘is’ (equative), hola ‘maybe’, hos ‘Let it be!’, hou ‘you are’,
10d. chunu ‘to touch’: cho ‘Touch!’, choyo ‘He touched’, chola ‘He may touch’
10e. dhunu ‘to wash’: dho ‘Wash!’, dhoyo ‘He washed’, dhola ‘He may wash’
10f. dinu ‘to give’: de ‘Give!’ linu ‘to take’: le ‘Take!’ (In some dialects)
10e. ho ‘to be’, cho ‘to touch!’, dho ‘to wash’, de ‘to give’
The shapes of roots thus derived can be verified againt Indo-Aryan cognates (See Turner 1931), eg:
11. Cognates of <ho> ‘be’:
*bhoti (Rigveda) hoti (Pali) bhomi, hoi (Prakrit)
hor (Syrian Romani) bhoNo (DoD, W Pahadi) bhoN (PaD, W Pahadi)
hoya (Bengali) hoeb (Maithili) hona (Hindi)
hovaN (Lahanda) hovu~ (Gujarati) hoNe (Marathi)
12. Cognates of <dho> ‘wash’:
dhAvati (Skt) dhopati (Pali) dhovai (Prakrit)
thovel (Eur Ramany) thovyu (Arm Romany) dhoNo (Kum)
dhoiba (Assam) dhoya (Bengali) dhoiba (Oriya)
dhona (Hindi) dhoNa (Pan) dhovu~ (Guj)
13. Cognates of <ro> ‘weep’
roditi (Sanskrit) rodati (Pali) rodai (Prakrit)
rovel (Eur Rom) roar (Syrian Rom) roya (Bengali)
rona (Hindi) roNa (Panjabi) rovan (Lahanda)
rovu~ (Guj)
These data show that Nepali <o>-ending roots have changed into <u>-ending ones in a nasalized environment.
However, between <cho> and <chu> forms (10d) comparative data reveal that the underlying root form is <chu> rather than <cho>. Here, <cho> is not basic, but derived due to analogy with other <o>-ending roots presented in (10b, c, e). The comparative Indo-Aryan data reveal this hypothesis, eg:
14. <chu> ‘touch’
Chupati (Sans) chupati (Pali) chuvai (Prakrit)
chuhNa (W Pah) chu~ya (Bengali) chui~ba (Oriya)
chuhuNa (Pan) chuna (Hindi) chuhaNu (Sindhi)
chuvu~ (Guj)
The tendence of raising the nasalized mid vowel is found even with the front vowel <e>:
15. <de> ‘give’
Deti (Pali) dei (Pali) deNu (W Pah)
Deba (Oriya) dena (Hindi) deNa (Pan)
Devu~ (Guj) Dena (Ben) deNa (Oriya) dena (Hindi) deNe (Marathi)
deNu~ (Guj) deNe (Marathi)
Conclusion:
The mathematical strategy of root derivation is general and more reliable.