09.5.2.16:36: ANDĚLÉ A DÉMONI

is the Czech translation of the title of the book in front of me. (It's not mine, and I'm not reading it. I've never read any knihu amerického spisovatele Dana Browna*.)

a is 'and'. démoni should be obvious.

Now can you guess what anděl means?

I was initially surprised that the word wasn't angel, but then I thought about it some more and guessed that the following happened:

- Czech speakers heard the word angel pronounced like 'ahn-jel' [andʒel].

- Czech had no 'j' [dʒ] sound

- Czech speakers substituted dě [ɟe] for foreign ge [dʒe]

-é and -i are two of out three nominative plural endings for masculine animates:

anděl 'angel' >andělé 'angels'

démon 'demon'> démoni 'demons'

ďábel '?'> ďáblové '?' (not ďábelové; the -e- drops out of the stem)

Can you guess what ďábel means? Hint: Think of the ' in ď as a little ⁱ.

Short's Teach Yourself Czech (1993: 80-81) tries to explain which ending goes with which animate:

-i "is used for most"

-ové "is chiefly used with 'international' words for professions"

ďábel is not a profession (not normally!), but it is an international word.

-ové "is also widely used to form the plural of one-syllable words, including many names of nationalities": e.g.,

Švéd > Švédové (but Wiktionary also includes Švédi!)

-é is for animates ending in:

-tel: učitel 'teacher' > učitelé 'teachers'

-(č)an: Kanaďan 'Canadian' (notice no -i- in Czech!) > Kanaďané

-at: demokrat 'democrat' > demokraté

plus a bunch of random animates: e.g.,

Žid 'Jew' > Židé 'Jews'

*Lit. 'book of the American writer Dan Brown'.

knihu is the accusative singular of kniha 'book'.

'of' is conveyed by three different masculine genitive singular endings:

the -ého of amerického 'of [the] American'

the -e of spisovatele 'of [the] writer'

the -a-s of Dana Browna 'of Dan Brown'

Another kniha Dana Browna is Šifra mistra Leonarda (Cipher of Master Leonardo) which also contains the masculine genitive singular ending -a.

(The -a of Šifra is a feminine nominative singular ending. Suffixes with different functions can be homophonous: e.g., -é is also a neuter nominative plural ending.)

09.5.2.1:32: ANDĚL

I've been reading a lot about Czech lately. Today I came across the word anděl and was surprised to learn what it means. Can you guess?

A hint: Czech ě often signifies that the preceding consonant sounds as if it is pronounced with a following y*: e.g., anděl is approximately 'ahn-dyel', not 'Ann-dell'.

(5.2.1:52: Another hint: an anděl is not a ďábel!)

*This is not quite accurate, but close enough to help you answer the question. Here is a more precise table of what ě signifies when preceded by various consonants:

ě symbolizes that the preceding consonant was originally palatalized: e.g.,

tě [ce] < *tʲe

pě [pje] < *pʲe

mě [mɲe] < *mʲe

ě currently doesn't follow consonants not in the above table. Other historically palatalized consonants either merged with their nonpalatalized counterparts or have their own symbols: e.g., *kʲe > če [tʃe] or ce [tse] (but not *kě).

Czech has both č and c from earlier *kʲ because it underwent two waves of palatalization with different outcomes. There seems to have been a chain shift:

First palatalization: *ke > *kʲe > *će [tɕe] > če [tʃe] (leaving a *će-slot open)

Second palatalization: *kai (unaffected by first palatalization because it had no *e) > *ke > *kʲe > *će (filling the now-open *će-slot) > cě [tsje] > ce (not če)

	Syllable subject to first palatalization	Syllable subject to second palatalization
Stage 1	*ke	*kai
Stage 2	*kʲe	*ke
Stage 3	*će	*kʲe
Stage 4	če	*će
Stage 5		Old Czech cě
Stage 6		Modern Czech ce

Mann's Czech Historical Grammar compares Old Czech rucě ([rutsje] > modern ruce [rutse]) 'hand' (loc. sg.) to Lithuanian rankai 'hand' (dat. sg.) which presumably preserves the earlier diphthong. The original velar reappears in modern Czech ruka (nom. sg.; cf. the Lith nom. sg. rañka) before a nonpalatal vowel. (Is the ending -i related to the Sanskrit loc. sg. ending -i?)

(Russian has a simpler paradigm for 'hand': the loc sg. of рука is just руке, not руце. But there are palatalized derived forms like ручка 'little hand'!)

Sanskrit also underwent two waves of palatalization which did not occur in Latin:

First palatalization: *ke > śa [ɕɐ]

e.g., śatam 'hundred' : Latin centum [kentum]

(cf. Czech sto with s- from an even earlier wave of palatalization than the ones I mentioned earlier)

Second palatalization: *kʷe > *ke > ca [cɐ] (not [tsa] as in Czech)

e.g, catur 'four' : Latin quattuor [kwattuor]

(cf. Czech čtyři)

Earlier kai has become ke [kee], but has not yet become ce [cee]. Sanskrit has developed a new kai from earlier kaai:

aai > ai > e [ee]

A similar vowel shift affected a(a)u:

aau > au > o [oo]

Last night, I asked,

Can you guess an English cognate of these words [e.g., Sanskrit aṣṭa] other than eight?

I gave four Sanskrit ordinal numbers as hints:

saptama 'seventh': cf. September

aṣṭama 'eighth': cf. October

navama 'ninth': cf. November

daśama 'tenth': cf. December

These four months were the seventh, eighth, ninth, and tenth months of the Roman calendar.

Other English cognates of Skt aṣṭa are here.

The -ber of these month names is cognate to Latin mensis, which in turn is cognate to moon. The Sanskrit cognate of these m-words is maa 'measure'. (The Sanskrit word for 'moon' is chandra, which is unfortunately better known as Chandra Levy's first name.)

The -m- in the month names has nothing to do with Skt -ma which means '-th' which may be cognate to -mus in Latin primus 'first' (also in English), septimus 'seventh', and decimus 'tenth'. (Latin octavus 'eighth' and nonus 'ninth' lack the -mus ending.)

Last week, I mentioned ṣaṣ 'six' and aṣṭa 'eight' as offhand examples of 'primary' retroflexes in Sanskrit that weren't derived from Sanskrit-internal rules changing s to ṣ. But upon closer analysis these 'primary' sounds are actually secondary from an Indo-European as opposed to Sanskrit perspective. We've seen that the retroflex ṣ-es in ṣaṣ 'six' are the products of pre-Sanskrit changes. The retroflex ṣ and ṭ in aṣṭa 'eight' are also due to such changes. Proto-Indo-European *okʸtoo had no retroflexes or even an s-like consonant until it underwent the following processes:

1. Shift of PIE palatalized (y-flavored) *kʸ to *sh (probably via an intermediate stage *ch):

*okʸtoo > *ochtoo > *oshtoo

2. Lowering of PIE *o(o) to a(a):

*oshtoo > *ashtaa

3. Shift of the cluster *sht (not permissible in Sanskrit) to ṣṭ:

*ashtaa > aṣṭaa

In later Sanskrit, the final vowel was shortened:

aṣṭaa > aṣṭa

Iranian and Slavic languages also have s(h) from PIE *kʸ in 'eight' but no retroflexes:

Avestan ashta

Persian هشت hasht (is the h- influenced by haft 'seven'?)

(Like Sanskrit, Iranian languages underwent the *o > a shift. But

Russian осемь osem'

Czech osm

Serbian осам osam

Here are a few hints from Sanskrit:

saptama 'seventh'

aṣṭama 'eighth'

navama 'ninth'

daśama 'tenth'

Do they ring any English bells?

And can you figure out the Sanskrit suffix meaning '-th'? Extra credit if you can think of an English word containing the cognate of that Sanskrit suffix.

There is no doubt that the Sanskrit word ṣaṣ 'six' is older than English six. Possibly over 3,000 years older. But is it closer to the Proto-Indo-European word for 'six', *s(w)eks (last seen here; see here for *-w-)? I have converted English six to phonetic notation [sɪks] for easier comparison:

If each word got a point for each consonant or vowel that was preserved, Sanskrit would get a score of zero and English would get a score of 3.5 out of 5 (or 4 if we ignore *-w- which was reconstructed to account for Armenian). English six is more archaic (more like PIE *s(w)eks) than Sanskrit ṣaṣ even though it is younger. How is this possible?

The ancestor of Sanskrit underwent four changes over three millennia ago:

1. *e lowered to a

2. Final *-s became retroflex -ṣ after *-k-

3. Final *-kṣ simplified to -ṣ

4. Initial *s- shifted to ṣ- to assimilate to final -ṣ

On the other hand, English only underwent a single change:

1. *e raised to i (the two vowels are similar*)

(but this didn't happen in other Germanic languages which retain e: e.g., German sechs, Swedish sex)

There is nothing preventing an ancient language from undergoing changes that a modern language didn't. Conversely, there is nothing preventing a modern language from preserving traits that an ancient language lost long ago. One cannot assume that age entails archaism.

Here is a biological analogy. Imagine that Proto-Indo-European is a woman with two children, Indo-Iranian and Germanic. Both children don't inherit her *w-trait (if any), but they do inherit four other traits (*s, *e, *k, *s). The granddaughter of Indo-Iranian, Sanskrit, has lost all four traits. (Granddaughters don't have to look like their grandmothers.) However, the distant descendant of Germanic, English, retains three of the four traits (*s, *k, *s) and some of the fourth (i is similar to *e).

One should not judge an entire language by a single word. On average, I would say that Sanskrit is more archaic (i.e., conservative) than English. Sanskrit retains the original Proto-Indo-European case system and genders, whereas most English speakers have no idea what they're missing - e.g., up to 24 forms of a noun and 72 forms of an adjective**.

Nonetheless, one cannot automatically assume that Sanskrit is always more archaic than English. For example, Dravidian influence has led to the development of Sanskrit retroflex consonants which are not retentions from Proto-Indo-European and which are generally absent in other Indo-European languages***.

*i and e are so similar that some American English dialects have merged them: e.g., pin and pen have become homophones.

**Not all of these forms are phonetically distinct: e.g., most masculine and neuter forms sound alike. But Sanskrit still has more forms than English: e.g., deva 'god' in Sanskrit has 17 phonetically different forms whereas English god only has two phonetically different forms ([gɔd] and [gɔdz]) since god's, gods, and gods' are all homophonous.

***Some varieties of Swedish developed retroflexes from r-consonant clusters long after Sanskrit developed its retroflexes. A shared trait (retroflexes) is not necessarily indicative of common ancestry. One cannot say that Swedish is more closely related to Sanskrit than to English on the basis of shared retroflexes. Swedish is in fact more closely related to English, and its retroflexes only coincidentally resemble those of Sanskrit.

09.4.28.23:23: KOREAN AS AN OFFICIAL LANGUAGE OF AUSTRALIA AND NEW ZEALAND?

Or should I say 오 스트레일리아 osUthureillia and 뉴질랜드 nyujillEndU?

In this context, a pact with Australia and New Zealand to use Korean and English as common official languages should be proposed as the best way to lay a solid foundation for concrete measures to achieve greater friendship and cooperation.

A language FTA [free trade arrangement - would Australia and New Zealand freely choose this?] would mean Korean, though geographically surrounded by the two mega-languages Chinese and Japanese, would overtake them in terms of practical use in Oceania.

Uh, there's more to Oceania than those two countries!

Korean would be taught at schools at all levels by Korean teachers with high proficiency in English.

How many such teachers are there, and how many of them would be willing to move to Australia and New Zealand?

Korean would be used along with English in all public documents and road signs.

Did he just say "all"!?

Korean residents there will be the initial beneficiaries of such an accord.

So let me get this straight - two countries are supposed to bend over backwards for 54,000 Korean speakers in Australia and 29,000 Korean speakers in New Zealand. Riiiight.

Conversely, more Australian and New Zealand teachers with high proficiency in Korean will teach English in Korea.

And how many of those people exist?

Relax, this is just one professor's proposal. But what a proposal!

Brian Deutsch and his readers offer more commentary.

09.4.28.22:36: DOES AGE EQUAL ARCHAISM?

While I was talking to David Boxenhorn last night, a question occurred to me:

Which language has the more archaic word for 'six', English or Sanskrit?

In other words, which word is closer to the Proto-Indo-European (PIE) word for 'six' - English six or Sanskrit ṣaṣ (pronounced shahsh)? Which word is more like its ancestor?

The oldest Sanskrit text, the Ṛgveda, is at least three millennia old, whereas you're reading English right now. The age of Proto-Indo-European is uncertain, but it must predate Sanskrit, and the interval between PIE and Sanskrit must be shorter than the interval between Sanskrit and English. So Sanskrit is closer to PIE in terms of time - but is the Sanskrit word ṣaṣ closer to the PIE word?

Ponder the title question until I post the answer tomorrow.

to Guillaume Jacques' superior knowledge. He pointed out that the retroflexion of ṣ- in ṣaṣ 'six' is not a sui generis phenomenon but can also be found in

viraaṣaaṭ 'subduing men' (nominative singular, Ṛgveda 1.035.06b; Monier-Williams has viraaṣaat with a typo -t) <

viira 'man; hero' (with irregular shortening of ii) +

why does -a lengthen in this compound (viira- > viiraa-) and in amaraavatii (< amara + -vatii)?

(Monier-Williams refers to Paaṇini 6.3.119 for the latter case but that's missing from this site!)

(p. 84 of Whitney's Sanskrit Grammar simply says "The final vowel of a former member of a compound is often made long, especially in the Veda.")

ṣaaṭ < ṣaaṣ < ṣaakṣ < *saakṣ < *saaks < *saagh-s 'overcome' (root sah with lenghtening < Proto-Indo-European *segh; English cognates here)

cf. 'six': ṣaṣ < *ṣakṣ < *sakṣ < *saks < *seks

in absolute final position, ṣaṣ 'six' becomes ṣaṭ

The process is the same in both 'subduing men' and 'six':

-. final -s becomes retroflex ṣ after k by a regular rule

- root-initial s- assimilates to the following retroflex ṣ: *sV(k)ṣ > ṣV(k)ṣ

(It's not clear when the -k- of -kṣ drops out.)

I thought the second part was irregular, but it may be regular though infrequent because there are not too many cases of *sV(k)ṣ in pre-Sanskrit. Whitney gives other examples of sah with double retroflexion:

ṣaaṭ ?'overcoming' (nom. sg.; sah as an independent adjective; p. 152)

turaaṣaaṭ 'overpowering the mighty' (nom. sg.) < tura 'mighty (p. 75; with inexplicable stem-final lengthening; this example added 4.28.0:54)

cf. turaaṣaaṭsu (loc. sg. with a retroflex-alveolar sequence -ṭs-; not -ṭṣu!); I would expect turaaṣaakṣu with -kṣ-, but I presume the nom. sg. -ṭ has been generalized as the stem to which -su was added

satraaṣaaṭ 'always overcoming' (nom. sg.) < satraa 'always' (p. 64)

cf. satraasaaham (acc. sg.) without retroflexion because there was no *-ks sequence to condition the first stage of retroflexion: the *-gh of the root simply lenited to -h- before the ending am

4.28.0:15: ADDENDUM 1: Whitney also gives -ṣaaham as the acc. sg. of -sah in compounds. I presume this ṣ is by analogy with the nom. sg. -ṣaaṭ since -saaham < *-saagham which has no *-ks sequence

4.28.1:04: ADDENDUM 2: sah is a root I love and hate.

Love because its forms are so unusual: e.g.,

soḍha ~ saaḍha 'subdued' as well as regular sahita < -(i)ta (past passive participle suffix)

cf. uuḍha ~ voḍha 'carried' < vah 'carry' (there is no udha < *wh-ta or vahita or vadha 'carried' < *vah-(i)ta; h is [ɦ])

saaḍhvaa in addition to regular sahitvaa 'overcoming' < sah + -(i)tvaa (gerund suffix)

Why is there retroflexion in forms of -h < *-gh roots which never had *-ks sequences? Is the retroflexion generalized from forms containing -ṭ?

And why do the root vowels of sah and vah either become long o [oo] which is usually from *au or lengthen? Past passive participles usually have maximally short forms of roots, but sah and vah have longer forms.

Whitney wrote on p. 75,

This is as if we had to assume as transition sound a sonant aspirate lingual [= retroflex] sibilant ẓh, with the euphonic effects of a lingual and of a sonant aspirate, itself disappearing under the law of the existing language which admits no sonant sibilant.

So perhaps

*sagh-ta > *saẓh-ta > *saẓḍha > *saaḍha (loss of ẓ with compensatory lengthening)

But how did that "transition sound" develop retroflexion? Perhaps

*gh > *źh > *ẓh

just as

*k > ś (which did not become retroflex ṣ unlike its voiced aspirate counterpart)

*g > ?*ź > j (= dź; fortition?)

These earlier velars were, strictly speaking, Proto-Indo-European palatovelars (*kʲ, *gʲ, *gʱʲ) which is why they became palatal fricatives *ś, *ź, *źh (> Skt ś, j, ḍh).

4.28.2:57: Why did root a become long o [oo]? Perhaps ẓh sometimes became *w, maybe through some intermediate stage like Czech ř, "post-alveolar ... with considerable friction" (Short 1993: 457; i.e., fricative-like - hence [dvorʒak] is an approximation of Dvořak):

*saẓḍha > *sařḍha > *sawḍha > soḍha

This sounds odd at first, but note how r can become w (e.g., in Elmer Fudd's pronunciation of rabbit as wabbit). Also note the secondary -w- from Late Middle Chinese *-ɨ- after Mandarin retroflexes: e.g.,

霜 'frost': *ʂɨaŋ > shuang [ʂwaŋ]

(but Sino-Vietnamese sương [ʂɨəŋ] without -w-)

*-ɨ- becomes nonlabial [j] after nonretroflexes:

相 'mutual': *sɨaŋ > *sjaŋ > xiang [ɕjaŋ]

sehe 'I subdued' (perfective) in addition to sasa(a)he with normal reduplication

The contraction of *Ca-CaC to CeC is actually normal, but I don't understand the mechanism, as -e- is usually from *ai, and there is no *i in *Ca-CaC.

Hate because the unusual is hard to learn.

4.28.2:44: ADDENDUM 3: Whitney (p. 64) lists yet other instances of s > ṣ after a that cannot be explained via kṣ: e.g.,

sa- 'with' + stubh 'utter joyful sound' > sa-ṣṭubh 'a kind of poetic meter'

apa- 'away' + sthaa 'stand' > apaaṣṭha 'barb of an arrow' (with inexplicable lengthening of the second vowel of apa-; the shortening of sthaa 'stand' is expected in compounds: e.g., yogastha 'absorbed in yoga' - not yogaaṣṭha!)

There's also an apaṣṭha 'point of a hook for driving an elephant' made up of the same elements but without the lengthening of the second vowel of apa-.

Presumably the meaning of bapaṣṭha is from '(something that makes an elephant) stand away', and 'barb of an arrow' resembled this hook.

09.4.26.23:59: RETROFLEX *SEKS

What will I stoop to next to boost blog readership? Vowel violence?

Seriously, *seks is the Proto-Indo-European word for 'six'* with two normal s-es. Its Sanskrit descendant ṣaṣ has two retroflex ṣ-es (note the subscript dots) and is one of the few basic words in Sanskrit with retroflexes (cf. Zulu whose basic vocabulary is full of clicks due to Khoisan influence). Why did s turn to ṣ in 'six'?

One might guess that the answer is Dravidian influence, but proto-Dravidian had no retroflex *ṣ (or an *s of any sort), and I think retroflex ṣ in modern Dravidian languages is in loanwords from Sanskrit. ṣ is the most common of the Sanskrit retroflexes even though it is not of Dravidian origin. Sanskrit s shifts to ṣ in native words in certain environments: e.g., after k:

*seks > *saks > *sakṣ

Final -kṣ was then reduced to retroflex -ṣ in 'six' (Whitney, Sanskrit Grammar, p. 51). Sanskrit does not permit final consonant clusters.

But what about the initial retroflex ṣ-? Perhaps it is due to irregular regressive assimilation. Anticipating a retroflex ṣ later in 'six', speakers pronounced a retroflex ṣ 'ahead of time' at the beginning of the word:

*sakṣ > *ṣakṣ

Normally pre-Sanskrit *se simply becomes sa with regular s-, not retroflex ṣ: e.g.,

PIE *sed > Skt sad 'sit' (cognate to sit; more cognates here)

PIE *septm̩ > Skt sapta 'seven' (cognate to seven; more cognates here)

4.27.00:40: One might think there's a connection between Sanskrit ṣ- and Slavic š- in 'six': e.g.,

Russian шесть

Czech šest

Serbian шест

(Why do they end in -t? Is that a remnant of a post-PIE suffix *-tĭ absent from other branches? Is it due to the influence of the preceding numeral ending in Proto-Slavic *-tĭ: Rus пять 'five', etc.?)

However, PIE *se did not normally become Slavic še-: e.g., PIE *septm̩ 'seven' did not develop initial š- in Slavic:

Russian семь (not шемь)

Czech sedm (not šedm)

Serbian sedam (not šedam)

4.27.0:43: If you want to see words for 'seven' in a lot of languages on a single page, go here. You can also find large numbers of translations for other words in Wiktionary. And if something isn't in Wiktionary, go to its Wikipedia entry, look for your target language on the left, and find the title of the article in that language. Wikipedia itself is an unintentional multilingual dictionary.

*4.27.0:34: The word has been reconstructed as *s(w)eks (Watkins) or *(s)weks (Beekes) with *w to account for the v- of Armenian vec' 'six'.

4.27.1:36: PIE *w- also can harden to g- in Armenian (Beekes 1995: 135). Cf. the hardening of *w- to [g] in Germanic loans in French: e.g., guerre corresponding to Eng war (< ult. PIE *wers; more cognates here). Other examples here. *w- (Latin v) became v- in native words: e.g.,voix 'voice' < Lat vox (cf. Skt vaac).