Worlds collide: On Thursday, I wrote about the southern African Sprachbund and mentioned the World Book Encyclopedia whose computer edition is produced by Software MacKiev which publishes The Print Shop, a program that once belonged to Brøderbund. On Friday, I saw a facebook post about The Warriors that also led to Brøderbund via Carmen Sandiego (don't ask). So I'm bound to write about the etymology of Brøderbund:

The word brøderbund is not an actual word in any language, but is a somewhat loose translation of 'band of brothers' into a mixture of Swedish, Danish and German. The 'ø' in Brøderbund was used as a play on the Danish letter ø and the slashed zero found in mainframes terminals and early personal computers.

The company's name was pronounced
BRU-der-bund instead of the publicly used BRO-der-bund, although in Danish, the vowel would sound roughly like the long u [sic! -it's ɚ] in hurt.

The Danish word for 'brother' is broder (no ø), and its plural is brødre with -re, not -er. (Cf. English brethren with -re corresponding to singular brother with -er. The first -e- of brethren is presumably from a fronted -o- corresponding to the front ø in brødre and the front ü in German Brüder [singular Bruder].)

The Swedish word for 'brother'is also broder (no ø - a letter absent from Swedish), and its plural is bröder with Swedish ö (a letter absent from Danish) instead of ø.

Thus brøder is like a Danish respelling of Swedish bröder 'brothers'.

Bund is much simpler. It's just German for 'union; band' and it's cognate to English band, bend, bind, and bond. More cognates here.

Both parts of Brøderbund have cognates in Sanskrit, so the name could be translated as



bhraatṛ is 'brother' and bandha (root bandh 'bind') is 'bond'. REWARDING THE WEIRD

Here's the ultra-condensed version of the last post for AWTM:

The Nguni languages of South Africa (e.g., Zulu) acquired an unusual feature (clicks) from their Khoisan neighbors but not other less exotic features. Why? Robert K. Herbert proposed that when Khoisan women married Nguni men and adopted the Nguni practice of hlonipha (linguistic taboo), they used the sounds they already knew - the clicks of their native languages - in substitutes for tabooed words because hlonipha encouraged maximum deformation. Clicks insured that replacement words sounded very different from the originals. Children of these mixed marriages would grow up with clicks and accept them as normal.

But there's one question that Herbert didn't ask: what did the Nguni husbands think of this? Were they really comfortable with hlonipha incorporating alien sounds? Did any of them think, 'Yes, we want you to avoiding saying certain things, but no, we don't want you to resort to the clicks of your original language'? Did they learn those sounds from their wives? Or did they simply not matter in the long run because their children acquired the sounds they rejected (or at least had trouble with)? THE LINGUISTIC UNION OF SOUTHERN AFRICA

(The Union of South Africa ceased to exist in 1961, but it's one of the many things that are stuck in my head because I grew up reading and rereading the 1957 edition of the World Book Encyclopedia.)

Both Indian and southern African languages are examples of Sprachbünde. A Sprachbund (language union)

is a group of languages that have become similar in some way because of geographical proximity and language contact. They may be genetically unrelated, or only distantly related. Where genetic affiliations are unclear [or nonexistent], the sprachbund characteristics might give a false appearance of relatedness.

Indo-Aryan languages like Sanskrit and its daughter languages (e.g., Hindi and Bengali) are not related to Dravidian languages like Tamil. Similarly, the southern Bantu (Nguni) languages (e.g., Zulu) are not related to the Khoisan languages (which may not even be related to each other!). Nonetheless, Sanskrit shares features such as the presence of retroflex consonants with Dravidian:

In a classic 1956 paper titled "India as a Linguistic Area", Murray Emeneau laid the groundwork for the general acceptance of the concept of a Sprachbund. In the paper, Emeneau observed that the subcontinent's Dravidian and Indo-Aryan languages shared a number of features [like retroflection] that were not inherited from a common source, but were areal features, the result of diffusion during sustained contact.

And Nguni languages share clicks with Khoisan languages:

The Nguni languages of Southern Africa, including Zulu and Xhosa evolved from the Bantu languages of the Congo area, which do not use clicks. During and after the Nguni migration to Southern Africa, the Nguni came into frequent contact with speakers of the Khoisan languages, which make abundant use of click sounds. Over time, the Nguni languages started to incorporate click sounds, until they became normal consonants as they are today.

But the degree of sharing differs as we saw in the last post. Retroflexes are marginal in Sanskrit, whereas clicks have penetrated Zulu basic vocabulary. And Zulu doesn't even have as many clicks as its relative Xhosa which has three more. (The Xh- of Xhosa is a click.)

Why are clicks so much more integral than retroflexes? David Boxenhorn speculated that this may reflect frequencies in Dravidian and Khoisan. Proto-Dravidian only had a few retroflexes (and Sanskrit has generalized the retroflexion feature to voiced stops and aspirates (ṭh, ḍ, ḍh) absent in PD), but Khoisan languages are rich in clicks. !Xóõ (! is a click) in Botswana and Namibia has 83 (!) different clicks as part of the largest consonant inventory of any known human language. These many clicks

are perfectly normal consonants in Taa [= !Xóõ], and indeed are preferred over non-clicks in word-initial position. (Emphasis mine.)

Other Khoisan languages also have impressive though lower numbers of clicks: e.g., ǂHõã is a click) has at least 55 clicks and !Kung (! is a click) and Juǀʼhoan is a click) have 48 each. If your neighbors clicked that much, you might start clicking too.

4.24.1:54: ADDENDUM 1: David Boxenhorn suggested that I compare Nguni languages to see if I can find correspondences between clicks. The Wikipedia article on Nguni languages provides a few examples which I've marked in bold:

'I only understand a little English.'

Ndi-qonda isi-Ngesi ka-ncinci nje. (Xhosa)

Ngi-qonda ka-ncane nje isi-Ngisi. (Zulu)

Gi-visisa si-Kguwa ka-nci të-jhë. (Phuthi)

Ngi-qonda ka-ncane nje isi-Ngisi. (Ndebele)

Ngi-condza ka-ncane nje si-Ngisi. (Swati)

(4.24: Zulu -d- corresponds to Swati -dz-, so is Zulu d partly from *dz, which might account for its high frequency? But then I would expect Zulu t to be partly from *ts, so t should also be very frequent, though it is rarer than d, at least in initial position.)

In Zulu, qonda is 'understand' and ncane is 'little'. (Ngisi (no click) is 'English'.) 'Understand' and 'little' have clicks q- and nc- in nearly all five languages. I don't know their etymologies. They could be

a. Khoisan loanwords (into Proto-Nguni?) - but they're basic vocabulary which is less likely to be borrowed (though anything can be borrowed!)

b. native Bantu words whose consonants have shifted to clicks under Khoisan influence (cf. the native Sanskrit words which have retroflexes such as niiḍa 'nest' < *ni-sda; sd is cognate to 'sit').

I can't find any obvious cognates at kamusiproject.org (kamusi < Arabic qamus 'dictionary'; but granted, Swahili is a very distant relative of the Nguni languages) or this online Proto-Bantu database or in some of these Khoisan databases (I haven't tried them all).

4.24.2:14: ADDENDUM 2: airforcewife drew my attention to this hypothesis about clicks and early human language.

4.24.2:54: ADDENDUM 3: According to Robert K. Herbert in "The sociohistory of clicks in Southern Bantu" (1995),

... click consonants [in Southern Bantu languages] are not reflexes of inherited elements; rather, the clicks were borrowed" [implying that words with clicks are generally borrowed - see below - though I imagine at least some onomatopoeia is of native origin and date after the borrowing of clicks] ...

It is frequentrly estimated that about 15 per cent of Xhosa and Zulu words exhibit clicks; the vast majority of these are words of demonstrable or presumed Khoesan origin, but there are examples where a click inexplicably substitutes for an inherited Bantu consonant. [Herbert later gives the examples of

Zulu xhopha 'hurt the eye' sharing a root with Zulu ukhophe 'eyelash'.

[Zulu] -cima 'extinguish' < Proto-Bantu *lima (though c is not even a lateral click!)

According to Lanham (1964), 21-25 of the 55 Xhosa consonants are non-inherited and confined almost exclusively to the borrowed vocabulary. [I presume those 21-25 consonants include the 18 clicks of Xhosa.]

Herbert points out that one would expect clicks to borrowed as nonclicks, but instead clicks were borrowed in huge numbers. Conversely, less exotic nonclick aspects of Khoisan phonology (nasalized vowels, diphthongs, and word-final consonants) were not borrowed.

Herbert thinks that

- Khoisan and Bantu speakers intermarried for centuries; 60% of the Xhosa and 49% of the Zulu share an allele with the Khoisan - but he also admits there is no correlation between genetic admixture and linguistic influence - "genes do not speak languages".

- the Nguni practice of hlonipha (linguistic taboo) facilitated the borrowing of clicks:

... the native (i.e. Khoesan) phonological inventories provided Khoe, San and Nguni women [who married Nguni males] with a ready-made and 'natural' source for consonant [and lexical?] substitutions as required by hlonipha ... a woman who enjoys a prohibition against uttering the syllables bo, nga, ni, di, ke, sa, etc. would look to this alternative phonetic inventory in order to replace Nguni consonants. Bear in mind here that the pre-contact Nguni consonant inventory was relatively small. The substitution of a foreign element such a click is perceptually salient and deforms the offending syllable acceptably [whereas less exotic Khoisan sounds might not be as deforming].

As I looked through Zulu words with initial clicks in Doke and Vilakazi's Zulu-English Dictionary, I noticed hlonipha forms - e.g., just now I found

-cakuva 'hlonipha term for umhlakuva, castor-oil bush'

with the click c as a substitute for hl (even though c is not lateral like hl!).

However, not all hlonipha words contain clicks - e.g.,

-amuleka 'receptacle; hlonipha term for isandla, hand'

(with the substitution of a phonologically totally different but loosely semantically related root)

so I was not sure if hlonipha was relevant to the borrowing of clicks.

Hlonipha is apparently becoming increasingly irrelevant in modern South Africa:

... many younger speakers of Xhosa in rural areas do not use traditional hlonipha at all, and they often use English or Afrikaans words when they attempt to do [so] ... 'many school children consider the whole concept a joke'. Similarly, many urban Zulu postgraduate students have described their reading of the hlonipha literature as 'like reading about a foreign culture'.

Are many of the hlonipha terms in my 1953 Zulu-English dictionary extinct 56 years later?

But I wonder if hlonipha is relevant to Tangut. Could the mysterious Tangut ritual language synonyms for normal native words have originated from an earlier Tangut version of hlonipha?

The whole of Language and Social History: Studies in South African Sociolinguistics looks interesting, but I'm out of time! I'm ordering a copy from Amazon for $18.78 plus shipping and handling. SANSKRIT RETROFLEXES VS. ZULU CLICKS

Sanskrit developed retroflexes under the influence of Dravidian languages. Similarly, southern Bantu languages like Zulu developed clicks under the influence of Khoisan languages. But how far do the parallels go?

Although retroflexes are commonly associated with Indian languages, they are actually relatively rare in Sanskrit, which is what we might expect for sounds that were originally non-native. Whitney's Sanskrit Grammar lists the frequencies of each segment in Sanskrit. Dentals are six times more common than their retroflex counterparts:

Sanskrit retroflex




% in text sample

0.26 0.06 0.21 0.03 1.03 1.45 3.0

% of corresponding dental

6.65 0.58 2.85 0.83 4.81 3.56 19.3
Dental-to-retroflex ratio 25.6 9.7 13.6 27.7 4.7 2.5 6.4

A random dh in Sanskrit is almost twenty-eight times more likely to be dental dh rather than retroflex ḍh.

Primary retroflexes are rare in Sanskrit basic vocabulary. Offhand, I can only think of ṣaṣ 'six' and aṣṭa 'eight'. Secondary retroflexes are conditioned and are not inherent in roots: e.g., n becomes when preceded by r in words like maraṇam 'death' < mara- + -na- with dental -n- + -m. Whitney's figures conflate primary and secondary retroflexes.

Whitney's Sanskrit Roots lists

- four roots with ṭ-; only one has more than a single attestation; the rest are dubious

- no roots with ṭh- or ṇ-

- four roots with ḍ-; two are dubious

- one root with ḍh-

- one root with ṣ- (ṣṭhiiv), the only retroflex-initial attested as far back as the Vedas

If dubious roots are excluded, only five are left and none are basic.

Zulu clicks are far more common than Sanskrit retroflexes. 11 out of 15 Zulu clicks appear in 21 items (10%) of a 207-word Swadesh list:

Zulu click
















Swadesh 207 words

think, lie (in a bed), squeeze, sing



small, narrow, suck, yellow


thick, leaf, egg, turn (v.i.), road





bark (of a tree)

Here are the number of pages for words with initial clicks in Doke and Vilakazi's Zulu-English Dictionary:

By individual click

Zulu click
















Number of pages
















By number of pages

Zulu clicks

c-, q-

ch-, nc-, qh-, gq-, nq-

gc-, ngq-, x-, xh-

ngc-, gx-, nx-, ngx-

Number of pages





By type of click

Click type


Aspirated: -h-

Voiced aspirated: g-

Nasalized: n-

Nasalized aspirated: ng-

Number of pages






Lateral clicks are generally the rarest. The page count for all laterals combined (22) is less than the page counts for dental c- and alveolar q-.

Compare the above page counts with those for Zulu alveolar nonclicks:

Zulu alveolar nonclick t- nt- th- d- nd- l- hl- nhl-
Number of pages 18 9 35 48 7 30 36 3

There is no Zulu nth- or nl-.

There are several surprises:

(n)c- and (n)q- are more frequent than their nonclick counterparts (n)t-.

gq- (but not gc-) is more frequent than its nonclick counterpart d-.

ngq- (but not ngc-) is about as frequent as its nonclick counterpart nd-.

th- is almost two times more common than t- (cf. Sanskrit in which t- is more than ten times as common as t- and Mandarin in which unaspirated d- [t] is slightly more common than aspirated t- [th])

d- is two and a half times more common than t- (cf. Sanskrit in which t- is more than twice as common as d-)

hl- [ɬ] is more common than l-

Could there be historical reasons for these anomalies: e.g., is d- a merger of several earlier initials? ZULU SPELLING III: C/Q/X

If I didn't already know what those three letters stood for in Zulu, I would guess that

c was 'ts' (as in some Eastern European languages) or 'ch' (which is actually spelled tsh which is not [tsh]!)

q was like Arabic q (which does not exist in Zulu)

x was 'sh' (which is actually spelled sh in Zulu)

All three letters turn out to represent clicks. Zulu has 15 clicks in all - all variations of c, q, and x written with or without added letters that should not necessarily be taken literally: e.g, gc- is not g + c.

Basic Aspirated: -h- Voiced aspirated: g- Nasalized: n- Nasalized aspirated: ng-
c ch gc nc ngc
q qh gq nq ngq
x xh gx nx ngx

Clicks are not an original Bantu feature. Southern Bantu languages with clicks like Zulu were influenced by the non-Bantu Khoisan languages.

Although people around the world make clicks, nearly all the languages that use clicks as regular consonants are in Africa. The sole exception is Damin, a special register of Lardil in Australia. Damin is not a full-fledged language:

Damin had a much more restricted and generic lexicon than everyday language. With only about 150 lexical roots, each word in Damin stood for several words of Lardil or Yangkaal. It had only two pronouns (n!a "me" and n!u "not me"), for example, compared to Lardil's nineteen, and had an antonymic prefix kuri- (tjitjuu "small", kuritjitjuu "large").

n! is equivalent to the Zulu prenasalized click nq. All but one Lardil click are nasal.

Next: A glimpse of Zulu click frequency. ZULU SPELLING II: G/K

Before answering my question from part I, let me supply a few examples of loanwords with bh- in Zulu. Can you guess what they mean?





The last one is really hard unless you know Dutch.

I would expect Zulu regular [g] and implosive [ɠ] to be written as gh and g by analogy with [b] and [ɓ] which are written as bh and b. However, [g] is g without an h: e.g., igalaji < garage. And [ɠ] is ... k, the same symbol used to write [k]! Thus one has to know that the infinitive prefix uku- is [uɠu] (not [uku]) but ibhayisikobho is [ibajisikobo] (not [ibajisiɠobo]). This ambiguity must be troubling for learners. Native speakers presumably have no trouble with the two values of k, just as English speakers have no trouble with the three values of th:

[θ] in thesis

[ð] in these

[t] in Thomas, Theresa

I would like Zulu orthography to be more consistent and less ambiguous:

IPA Current orthography My hypothetical orthography
b bh b
ɓ b bh
g g g
ɠ k gh
k k

h would differentiate implosive [ɓ ɠ] from nonimplosive [b g] and k would always simply be [k]. (h has other functions in real Zulu orthography that I won't go into here.)

The sources of the four bh-words at the start of this post are

iBayibeli < Eng Bible (the iB- reminds me of i-consonant words like iPod)

ibayisekili < Eng bicycle

ubhanana < Eng banana (note u- instead of i-)

ibhayisikobho < Afrikaans bioscoop < Dutch bioscoop 'movie theater'

the hypothetical English equivalent would be bioscope

Next: Guess what Zulu c, q, and x stand for. ZULU SPELLING I: B(H)

As I wrote "South African Translation Strategies", I was puzzled by the -Bh- in Zulu isiBhunu 'Afrikaans' from Boer. Why would foreign b be borrowed as bh? (The substitution of -nu for -r reflects how Zulu lacks -r and doesn't allow words to end in consonants.) It turns out that in Zulu orthography, b is an implosive [ɓ] and Zulu bh is a regular [b]. Hence Zulu bh is not like a Sanskrit bh which is a voiced aspirate [bɦ]. (Nor is it like Irish bh [v] ~ [w].) I wonder why b and bh were not assigned to [b] and [ɓ] instead of the other way around. Could frequency be the reason? (But my Zulu-English dictionary has almost the same number of pages for b- and bh-: 42 and 41. Perhaps [ɓ] occurs in more frequent words and/or is more frequent in medial position.)

Zulu has regular [g] and implosive [ɠ]. Can you guess how they're spelled? Answer next time. TARZANCA*

- pronounced 'tarzan-jah' - is "a mixture of Turkish and English":

Even [on Turkish] television, some [Turkish] speakers announce an art exhibition in Tarzanca. Instead of başlamak [< baş 'head'] - 'to start' - they use start almak [lit. 'take'], which is neither Turkish nor English. Also, younger Turks commonly use English expressions, such as part-time, fulltime, prime time, art, cool, etc.

Tarzanca is the sort of hybrid that Setswana purists don't want. But would a 'Setarzan' be so bad?

The normal Korean verb for 'start' is 始作하다 shijakhada 'start', a Chinese-native hybrid not unlike Tarzanca start almak. (Shi- + -jak is 'start + make' - a Chinese compound made in Korea, or an extinct Chinese word? - and hada is 'do'.) Korean even has an English-native hybrid sUthathU hada, lit. 'start do'. Korean has lost its native words for 'river' (now 江 kang from Chinese) and 'mountain' (now 山 san, also from Chinese). Yet Korean is in no danger of extinction.

English itself is full of borrowings and continues to borrow.

*4.20.3:08: -ca is a suffix in Turkish names of languages: e.g., Almanca 'German language'. The root Alman 'German' is related to Spanish alemán and French allemand.

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2009 Amritavision