12.5.19.23:52: BUCKNELL'S SANSKRIT MANUAL 4: MAHIṢṬHAS TRILIṄGAKĀNĀM (logonote*)

Bucknell's Sanskrit Manual has twenty-two irregular masculine declensions. That's more than the number of regular declensions for any of the three genders. I'm not going to go through them all here, so don't run away from this blog just yet.

Perhaps the most important of the twenty-two is the last one:

62. mahānt- < *meǵʕ-ont-s 'great' (the first of the twenty-two is numbered 41)

I mention it not because it's interesting - like its neuter counterpart mahat it has irregular lengthening** - but because I happened to discover the University of Texas at Austin's Indo-European Lexicon when looking for its appearances in the Rigveda. Here's the Lexicon's entry for the Proto-Indo-European root of mahānt- and its cognates in many European languages (but not Dutch: e.g., meester < Latin magister 'master'). I used to link to the online edition of Watkins' (2000) The American Heritage Dictionary of Indo-European Roots at bartleby.com but that was taken down years ago. Now I finally have a substitute that goes beyond it by even including names and words from Tolkien (though none are in this particular entry).

*5.20.2:55: mahiṣṭhas 'greatest' is the superlative of mahānt-. The ending -iṣṭhas is from *-is-tHo-s with an unknown laryngeal *-H- and is cognate to English -est.

The comparative is mahīyān < *-iōs with an ō-grade version of the *-is- suffix of the superlative. (Perhaps the -n is due to confusion with masc. nom. sg. *-n elsewhere.)

Another comparative suffix is -tara which vaguely reminds me of English -er but probably isn't cognate, as English has no -th- corresponding to Skt -t-. There is no -tara comparative for mah- 'great'.

triliṅgakāṇām 'of adjectives' is the genitive plural of triliṅgaka- 'adjective' (lit. 'three-gender' since adjectives can have any of the three genders). Similar words for 'adjective' are

anyaliṅgaka- lit. 'other-gender' (i.e., a word that takes on the gender of another word: i.e., the noun it modifies)

sarvaliṅga- lit. 'all-gender'

So the title means 'greatest of the adjectives'.

**5.20.3:18: The irregular lengthening is in the 'strong cases':

I prefer to reflect this lengthening in my citation form mahānt-, but UT Austin prefers mahant- and Monier-Williams has mahat- sans -n-.

Cf. the declension of a regular participle which has no lengthening: 12. nayant- (never *nayānt-)

The distribution of long ā is the same as in declensions 9-10 for -an stems, 14 for the comparative (below), and 15-16 for perfect active particples.

Perhaps the long ā of mahānt- was influenced not only by those declensions but also by its unusual combining form mahā- (instead of the expected maha-). Long vowels were once common in combining forms but were mostly lost in later Sanskrit. Maybe the long -ā- of combining forms with a-stems was from *o with lengthening due to Brugmann's law; cf. the -o still in Slavic combining forms today. Non-a long vowels in combining forms could be due to analogy with the large class of a-stems whose combining forms ended in long -ā-.

12.5.18.23:59: BUCKNELL'S SANSKRIT MANUAL 3: MAHĀ MATA HARI (logonote*)

There are only three irregular neuter declensions in Bucknell's Sanskrit Manual:

63. mahat < *meǵ-ʕ-n̥t 'great' (cognate to mega-, magni-, much; ending cognate to Latin -nt)

declines almost exactly like a participle except for the long ā in the nominative / accusative / vocative plural: mahānti (instead of the expected *mahanti); lit. 'being great' < √mah 'great'

its feminine mahat-ī, like all (?) feminine adjectives, is regular

64. akṣi 'eye' (cognate to op-, ocul-, eye), asthi 'bone' (cognate to osteo-, os-), dadhi 'yogurt', and sakthi 'thigh'

have paradigms mixing the 18 -i and 27 -an paradigms (though -an normally alternates with -a, not -i!): -i before zero and consonantal suffixes, -n before vowels:

	singular	dual	plural
nominative accusative vocative	-i stem + -Ø, -nī, -ni following lengthened ī (voc. sg. optionally -e instead of -i)
instrumental	-an stem + -ā, -e, -as, -i	-i stem + -bhyām, -bhis, -bhyas
dative
ablative
genitive		-an stem + -os	-an stem + -ām
locative			-i stem + -su

65. ahar 'day'

has a paradigm mixing -ar, 23 -as, and 27 -an stems:

	singular	dual	plural
nominative accusative vocative	-ar stem + -Ø	-an stem + -ī, -i following lengthened ā
instrumental	-an stem +  -ā, -e, -as, -i	-as stem + -bhyām, -bhis, -bhyas
dative
ablative
genitive		-an stem + -os	-an stem + -ām
locative			-as stem + -su

-r and -s merge into in certain environments, so a mixture of -ar (unique to this noun?) and -as is understandable, but -an is not, as n does not alternate or merge with r or s)

5.19.00:59: ūdhar 'udder' (cognate to udder) used to belong to this category but was later regularized as an -as stem

5.19.1:59: I am surprised there aren't more after realizing that I forgot about the an-alternating stems in Whitney (1896: 160):

Last month I wrote at length about the irregular paradigm of hṛd 'heart' (cognate with cardi-, cord-, cred-, heart) which alternates with a longer stem hṛd-aya-. According to my understanding of Whitney (1896: 149), other short-long alternators are

Note that this set of alternators sometimes has the short and long forms in the 'opposite' places relative to the previous set: e.g.,

nominative singular

asṛj 'blood' (short; not long asa < *asn̥, reduction of asan)

hṛdaya-m 'heart' (long; not short hṛd!)

genitive singular

asan-as 'blood' (long) or asṛj-as (short)

hṛd-as 'heart' (short) or hṛdaya-sya 'heart' (long)

The logic behind these two opposite patterns eludes me.

*5.19.2:02: The title refers to three of the irregular neuter declensions:

mahā- is the compounding form of mahat 'great' (n.)

Indonesian mata hari 'sun' < lit. 'eye (of the) day' refers to akṣi 'eye' and ahar 'day'

12.5.17.23:57: BUCKNELL'S SANSKRIT MANUAL 2: THE SEVEN IRREGULAR SISTERS OF WATER

One simple feature that distinguishes Bucknell's Sanskrit Manual from every other Sanskrit grammar I've ever used is its comprehensive list of 73 nominal declension types:

Regular

16 masculine

15 neuter

9 feminine

Irregular

22 masculine

3 neuter

7 feminine

One can quickly scan that ten-page list to see which types one hasn't learned yet. Looking at the table of contents of a conventional grammar isn't as effective since not every type is listed: e.g., only 19 in Whitney's table of contents. (All of Whitney is in Wikisource!)

Whitney regarded the irregular feminine noun ap- 'water' as subtype A of his fifth declension, whereas Bucknell labeled it as the only member of declension 71. Here are the other irregular feminine nouns (with my preferred citation forms):

66. jarā 'old age' (cognate to geriatric and possibly Zarathushtra > Zoroaster)

looks like it should belong to the regular 32 -ā category (the first feminine declension I ever learned) but is a hybrid of 32 and the ambigender 8 -ās category

5.18.0:34: The mixture of 32 and 8 goes back to the Vedas and continued "through the whole history of the language" (Whitney 1896: 155).

67. strī 'woman' (the term for feminine grammatical gender is strīliṅga*)

looks like it should belong to the regular 33 -ī category (the second feminine declension I ever learned), but 33 is for polysyllabic words, not monosyllables

68. lakṣmīs 'fortune', tarīs 'boat' ('that which crosses' < √tṝ 'cross'), tantrīs 'string' (< √tan 'stretch > Tantra, tantric)

look like they should belong to the regular 34 -īs category, but 34 is for monosyllabic words, not polysyllables

Whitney listed a fourth member of this class: tandrīs 'laziness'

69. dyaus 'sky' (cognate to Deus and Zeus) and

70. gaus 'cow' (cognate to cow, bovine)

look like they should belong to the regular 39 -aus category which only has two members, naus 'ship' (cognate to navy) and the rare masculine (!) word glaus 'ball', but have unusual vowel alternations (table added 5.18.00:50):

category	39	69	70
nominative singular (identical endings)	naus < *nāu-s	dyaus < *dyāu-s	gaus < *gāu-s
accusative singular (different stem vowels)	nāvam < nāu-am	divam (not *dyāvam < dyāu-am	gām (not *gāvam < *gāu-am)

(71. ap- 'water'; see "Ancestors of the Offspring of Waters")

72. gīr 'voice' and

73. pūr 'city' (as in Singapore 'lion-city') and dhūr 'yoke'

have the same paradigm apart from their vowels

Whitney listed more: stīr; jūr, tūr, dhūr, pūr, mūr, stūr, sphūr, psūr

I'd list āśīs 'blessing' as its own category 74 (rather than Bucknell's 72) because its stem ends in -s, not -r, and I'd move 72 and 73 into the regular feminines since I don't know of any -īr or -ūr nouns that decline differently

5.18.1:10: Whitney (1896:148) wrote:

The native grammar sets up a class of quasi-radical stems like jigamis desiring to go, made from the desiderative conjugation-stem (1027), and prescribes for it a declension like that of āçís [= my āśīs] thus, jigamīs, jigamiṣā, jigamīrbhis, jigamīḥṣu, etc. Such a class appears to be a mere figment of the grammarians, since no example of it has been found quotable from the literature, either earlier or later, and since there is, in fact, no more a desiderative stem jigamis than a causative stem gamay [the actual desiderative stem is ji-gam-iṣa- and the actual causative stem is gam-aya-; their shared root is √gam 'go'].

If such a class of desiderative nominals existed, then I might move my 74 into the regular feminines.

I wasn't even aware of 66 and 72 (= my 74) āśīs.

*5.18:00:40: It's fitting that the term for masculine grammatical gender also has an irregular noun: puṃliṅga < category 55 puṃs 'man'.

puṃs is also in the term for neuter grammatical gender: napuṃsaka < na-puṃs-a 'non-man'.

12.5.16.23:59: ANCESTORS OF THE OFFSPRING OF WATERS

Last night I was frustrated by my inability to identify how Avestan "Napá" in Wikipedia was related to Apąm Napāt. I had never seen á in Avestan romanization before, and I assumed it was a substitute for long ā. It took me over three hours to figure out that it was actually a substitute for 𐬃 å [ɔː] from *-ās, the long counterpart of ō from *-as. I think the rounding reflects an earlier retroflex *-ẓ [ʐ]:

*-ās [aːs] > *-āẓ [aːʐ] > *[aːɰ] > *-āw [aːw] > -å [ɔː]

*-as [ɐs] > *-aẓ [ɐʐ] > *[ɐɰ] > *-aw [ɐw] > -ō [oː]

The shift of retroflex ẓ [ʐ] to labial w has parallels in the development of secondary labials from earlier Chinese retroflexes: e.g.,

雙 Late Middle Chinese *ʂaŋ > Mandarin shuang [ʂwaŋ] 'double'

書 Early Mandarin *ʂu > Xi'an Mandarin fu 'book'

In this scenario, the different heights of -å [ɔː] and -ō [oː] reflect the different heights of long and short a. I am projecting Sanskrit's lower long ā / higher short a distinction back into Proto-Indo-Iranian. If Avestan short a was still higher than [a], it couldn't be as high as Avestan ə.

The Avestan letter 𐬃 <å> is a blend of 𐬆 <ə> and 𐬃 <ā> (Avestan is written from right to left). The schwa indicates that <å> [ɔː] was higher than <ā> [aː] (and reminds me of how Koreanㅓ <ŏ> has been transcribed as both [ə] and [ɔ]).

*-ās and *-as underwent somewhat different developments in Sanskrit before word-initial voiced segments:

*-ās [aːs] > *-āẓ [aːʐ] > *[aːɰ] > -ā [aː]

*-as [ɐs] > *-aẓ [ɐʐ] > *[ɐɰ] > *-aw [ɐw] > -ō [oː] but -a before vowels other than a-

Given all of the above, Avestan Napå in Apąm Napå is from *napās, which I presume is from *napāt-s with the masculine nominative singular suffix *-s.

(5.17.00:31: Its Latin cognate nepos < *nepot-s 'grandson, nephew, descendant' has a similar reduction of final *-t-s.)

Jackson (1892: 49) listed the formula

Orig. t + s = Skt. s (through intermediate ss §§185, 186).

but he gave no examples of final *-t-s and I still can't find the nominative singular napå (though the stem is on p. 58).

I think Avestan zam- 'earth' and ziiam- 'winter' also underwent *-āC-s > *-ās simplification (final forms from Skjærvø 2003: 54 with ii rewritten as y):

*zām-s > *zā-s > zå

(cognate to the second half of Russian Novaya Zemlya)

(Jackson 1892: 93 derived this from *zm̥̄-s with a long syllabic m.)

*zyām-s > *zyā-s > zå

(I'm surprised by -ām- given its cognates Skt himā and Russian zima.)

Sanskrit has different simplifications:

*napāt-s > napāt 'grandson' (5.17.00:08: all 15 attestations in Rigveda here)

*kṣām-s > kṣās 'earth'

*praśām-s > *praśān-s > *praśān 'quieting'

(5.17.00:06: This word may not date back to pre-Sanskrit, but I think this change might have occurred to other *-ām-s words.)

Note, however, that not all final clusters were reduced to single consonants in Avestan: e.g.,

*āp-s > āf-š (not *ā-s or -å) 'water'

corresponding to a hypothetical Sanskrit *ap < *ap-s. Not all forms of Skt ap are attested. Whitney (1896: 148) listed only nine out of twenty-four possible forms:

Note the dissimilation of -p to -d before -bh- in the instrumental and dative-ablative. No such dissimilation occurred in the Avestan dative plural aiβyō < *ab-bhyas (Jackson 1892: 84). Nor does it occur in other Sanskrit -p stems: e.g.,

dharma-gub-bhyas (not *dharma-gud-bhyas) (dat.-abl. pl. of dharma-gup 'guardian of law')

The varying lengths of the first vowel of Sanskrit āp-/ap- is correlated with accentuation:

accented ā in the stem is long:

ā́pas (nom. pl.)

unaccented a in the stem is short:

apás (acc. pl.)

(5.17.00:18: But one should not conclude that accentuation and length are always correlated: e.g., apás (acc. pl.) and nápāt both have accented short a, even though the latter also has a long ā that one might think would 'attract' the accent.)

Although ap- should normally go back to Proto-Indo-European *H-p- (*H- = uncertain laryngeal), I wonder if it is from an irregular variant *ʕ-p- of the root *ʕ-kʷ- of Latin aqua.

*Sanskrit avoids -r r- sequences: e.g.,

-ir + r- > -ī r- (-r loss with compensatory lengthening)

-īr + r- > -ī r- (-r loss; ī already long so no compensatory lengthening needed)

12.5.15.23:59: NEPHEW NEPTUNE?

Until I found this 2009 article by Gordon Whittaker last night, I had never noticed the similarity between Sanskrit Apām Napāt / Avestan Apąm Napāt 'Offspring of Waters' and Neptune (< Latin Neptūnus). I just found that the connection was first proposed by Georges Dumézil:

Dumézil though remarked words deriving from root *nep- [i.e., cognates of English nephew] are not attested in IE languages other than Vedic and Avestan [but what about Old Irish in the next sentence?]. He proposed an etymology that brings together [Latin] Neptunus with Vedic and Avestan theonyms Apam Napat, Apam Napá [not Napát with -t?*] and Old Irish theonym Nechtan, all meaning descendant of the waters [does Nechtan have that meaning in OI?]. By using the comparative approach the Indo-Iranian, Avestan and Irish figures would show common features with the Roman historicised legends about Neptune. Dumézil thence proposed to derive the nouns from IE root *nepot-, descendant, sister's son**.

Wikipedia lists several other proposed etymologies for Neptūnus.

The sound correspondences seem to work. OI cht is from Proto-Indo-European *pt: e.g.,

secht 'seven' < PIE septm̥.

However, I don't know if Latin -ūnus and OI -an can be accounted for.

I wouldn't go as far as Whittaker, who also links these words to a name in a completely non-Indo-European language: Sumerian Nudimmud*** and its variants, none of which have anything corresponding to -p-:

Nutemud (oldest attested form?)

Nutememud

Nudamud

Nadimmud ("artificially differentiated" - so is the a artificial?)

And even if the first half of Nudimmud was cognate to Napāt, where would the second half come from?

*5.16.1:12: As far as I know, Napāt shouldn't lose final -t in Avestan. I can't find any reference to such a loss in Jackson or Skjærvø's grammars. Nor would I expect one given that final -t is stable in the Sanskrit -t declension: there is no Sanskrit *Napā.

5.16.2:59: Ah, I see: the nominative singular of Avestan Napāt is Napå (Skjærvø 2003: 66):

The stem napāt- has the nom. from an h-stem napah-.

The Old Persian cognate of Avestan Napāt is napa without -t.

**5.16.1:32: Was *nepōt that specific? Both Beekes (1995) and Watkins (2000) defined it as 'nephew' which is ambiguous.

***5.16.00:55: Wikipedia has a native etymology for Nudimmud: nu 'likeness' + dim mud 'make bear'.

12.5.14.23:59: BUCKNELL'S SANSKRIT MANUAL 1: NEPHEWS OR GRANDSONS?

Today I got Roderick S. Bucknell's (1994) Sanskrit Manual in the mail. Although getting another Sanskrit grammar felt redundant at first, Bucknell's approach has many unique features that set it apart from the competition. I may highlight some of them in future entries.

For now, I only want to mention that the first thing I looked up was his treatment of -ṛ stems like mātṛ 'mother' which I blogged about yesterday. He described the masculine and feminine -ṛ stems on separate pages. I prefer to have both of them on the same pages so that they can be learned simultaneously. Moreover, he referred to them by their nominative singular as -ā nouns which can easily get them confused with true -ā stems and -an nouns whose nominative singulars also end in -ā:

A partial comparison of -ṛ, -ā, and -an stems

Bucknell chose mātṛ 'mother' as the model for feminine -ṛ stems like duhitṛ 'daughter'; it is arguably more basic than svasṛ 'sister' which he treated as a variant. Similarly, he chose pitṛ 'father' as the model male -ṛ stem kinship term and treated naptṛ (which I glossed last night as 'grandson') as a variant. Questions:

1. Why did he gloss naptṛ as 'nephew' rather than 'grandson'? I have not found any Sanskrit dictionary defining it as 'nephew', though the word is certainly cognate to nephew, nepotism, etc.

2. Did Sanskrit lose the meaning 'nephew'? The word means both 'nephew' and 'grandson' in Old English and Latin, and Watkins (2000: 58) reconstructed its Indo-European source *nepōt with both meanings.

3. Can one generalize and say *nepōt originally meant 'secondary younger male relative' (with sons being primary younger male relatives) which then narrowed in semantic scope, becoming 'grandson' in later Sanskrit and 'nephew' in English?

(5.15.00:59: Wiktionary defined Old English nefa as 'stepson' in addition to 'nephew' and 'grandson'. 'Stepson' also fits my proposed generic meaning 'secondary younger male relative'.)

4. According to Monier-Williams, naptṛ has a variant stem napāt- "only in the strong cases and earlier lang." So is this how the paradigm changed through time?

(5.15.00:54: The Sanskrit Heritage site gives a full paradigm for napāt without any forms of naptṛ.)

5. Was the later naptṛ - the equivalent of a hypothetical English *nephther - by analogy with other kinship terms?

6. 5.15.1:00: Did that analogy occur at the Proto-Indo-Iranian level? Monier-Williams listed both types of forms in Avestan: napāt and naptar. So my stage 1 might have been pre-Proto-Indo-Iranian and both Sanskrit and Avestan might have inherited the stems from stage 2. My inability to find napāt in some dictionaries and grammars leads me to think that naptṛ was the sole survivor in Sanskrit by stage 3.

12.5.13.23:59: MĀTUR DAY

This photo of a "Mather's (sic) Day" cake made me think of Sanskrit mātṛ 'mother' which contains the syllabic ṛ I've been blogging about lately. mātṛ has a unique declension that is like a compromise between those of its fellow feminine svasṛ 'sister' and its opposite-sex counterpart pitṛ 'father':

Why does 'mother' decline like 'father' in the first three instances? At first I thought it was analogy: i.e., that 'mother' used to decline like feminine 'sister' but mostly switched to the 'father' declension since 'mother' and 'father' share the semantic category of 'parents'. However, now I think both 'mother' and 'sister' might be regular feminine -ṛ declension nouns. Judging from Sanskrit alone, the long vowel before r in 'sister' but not in 'mother' could reflect a laryngeal *H in 'sister' absent from 'mother':

*swesoHr- > svasār- (short vowel *o + laryngeal *H = long ā)

*meʕter- > mātar- (short vowel *e + laryngeal *ʕ = long ā)

Beekes' (1995: 38) Proto-Indo-European *suésōr could be from *swesoHr-.

(5.14.1:28: Burrow 1955 regarded the long ā of 'sister' and -tṛ agent nouns as being "introduced from the analogy of the nom. sg." which ends in ā: e.g.,

svasā (nom. sg.) : svasār- (stem of acc. sg., dual. nom./acc./voc., nom. pl.)

Macdonell notes that the non-agent -tṛ agent noun naptṛ 'grandson' also has ā where svasṛ does: e.g., acc. sg. naptār-am. But the nom. sgs. of 'mother', 'father', and bhrātṛ 'brother' also end in -ā:

mātā (nom. sg.) but mātar-, not *mātār- (stem of acc. sg., dual. nom./acc./voc., nom. pl.)

pitā but pitar-, not *pitār-

bhrātā but bhrātar-, not *bhrātār-

So why were they immune from analogy? Higher frequency than 'sister', etc.?)

The different plural accusative endings (f. -ṝ-s, m. -ṝ-n) are expected since -n is unique to masculines.

-as is a common accusative plural ending for both masculine and feminine nouns, so it is not surprising that it later spread to 'mother' from 'father', making their endings identical.

The -u- in the shared genitive singular - though normal for the -ṛ declension - is unusual because it's not part of the normal gradation of syllabic -ṛ. Where did it come from? Beekes (1995: 176) reconstructed the Proto-Indo-European genitive singular of 'mother' as *mḗ-tr-s (presumably < *meʕtr-s), which I would expect to develop into Sanskrit *mā-tṛ-s, not mātur (< *mātur-s?).

(5.14.1:36: Burrow 1955 derived -ur from *-ṛš; cf. Avestan -ərəš, -arš.)

The -u- in mātur reminds me of the irregular -u- in the weak present stem of class VIII kṛ 'do' e.g., kur-v-anti 'they do' instead of *kṛ-v-anti (cf. its earlier class V* form kṛ-ṇv-anti 'id.'.) However, this kur- is from *kʷr- with the labial quality of *kʷ having become u. (I thought I came up with this myself today, but now I see that I might have first read about this solution in Burrow 1955.) However, the -sur and -tur genitive singulars cannot be from *sʷr or *tʷr since Proto-Indo-European did not have any labialized alveolars or dentals.

*5.14.1:26: Class VIII verbs have -v- before the third person present indicative ending -anti whereas Class V verbs have -nv- (here, -ṇv- with a retroflex -ṇ- that assimilated to the preceding ṛ).