Content uploaded by Pierre Bancel
Author content
All content in this area was uploaded by Pierre Bancel on Apr 21, 2020
Content may be subject to copyright.
Brave new words
Pierre J. Bancel & Alain Matthey de l’Etang
Association d’Études linguistiques et anthropologiques préhistoriques, Paris*
Contrary to the received idea that globally spread papa/mama words are
constantly reinvented by children in dierent languages, we show here
that these words are always inherited from the most ancient stages of their
respective families, with the exception of a number of borrowings – which are
not innovations, either. We then show that probabilistic calculations aiming to
demonstrate that global and other remote etymologies might be mere chance
resemblances are invalid, and that chance cannot be reasonably invoked in the
cases these calculations deal with. Consequently, the global convergence of
papa/mama words can only be a trace of a common heritage of all human
languages. Finally, we link this nding with others, indicating that these words
must have appeared early, most probably at the very origin of articulate language.
. e Proto-Sapiens kinship terms papa, mama and kaka
Our central claim is that most modern papa/mama words, so widespread in all lan-
guage families worldwide, may be traced back to a common origin. We use the name
Proto-Sapiens for the original ancestor language from which they have been inherited,
and which must have been the ancestral language of all known languages spoken by
modern human beings, who together constitute the species Homo sapiens.
On archeological and genetic grounds, Proto-Sapiens may be dated between
200,000 years ago (the approximate earliest date at which our species emerged in
Africa; McDougall et al. 2005; White et al. 2003) and 50,000 years ago (the approxi-
mate latest date at which our direct ancestors may have le the African continent and
began to spread their bones, genes, artifacts, and language over the rest of the world).
However, recent archeological ndings from several South African sites – Klasies
River Mouth (Deacon 2001; Singer & Wymer 1982), Blombos Cave (d’Errico etal.
2005; Henshilwood & d’Errico 2005; Henshilwood et al. 2001) and Pinnacle Point
* Mail should be sent to first author: pierrejbancel@hotmail.com. anks to Peter MacNeilage,
Claire Lefebvre, and John Bengtson for their useful remarks and corrections, and to Shahar
Fineberg and Zofia Laubitz for their fine proofreading job. Errors, of course, are solely ours.
Pierre J. Bancel & Alain Matthey de l’Etang
(Marean et al. 2007) – have revolutionized the dating of modern Sapiens anatomy and
behavior. Until recently, modern behavior was widely believed to have appeared no
earlier than some 40,000 years before present (yBP). But all these sites have revealed
numerous unambiguous traces of modern behavior (use of marine food, cooking food
on hearths, microlithic tools, polished bone tools, personal ornaments, geometrical
engravings, etc.) older than 80,000yBP, and up to 164,000yBP at Pinnacle Point.
Genetics-based datings like those of the “mitochondrial Eve” around 200,000yBP
(Cann et al. 1987), or the split of Khoisan people between 70,000 to 90,000yBP (Knight
et al. 2003), as well as the antiquity of the rst human occupation of Australia and New
Guinea (at least 46,000yBP, and perhaps 60,000yBP for Australia; see Bowler etal.
2003; O’Connell & Allen 2004) also tend to push Proto-Sapiens back to a date earlier
than 50,000yBP, perhaps as far back as some 100,000yBP.
. Historical background
e global distribution of papa/mama words, noted as early as the mid-nineteenth
century (Buschmann 1852), received its currently accepted explanation in the late
1950s. Murdock (1959) and Jakobson (1960) – probably drawing on Lubbock (1889)
or Westermarck (1891), though they do not quote any predecessor in this regard –
explained that modern papa/mama words must be recent and had resulted from
constrained, convergent innovations due to child/parent interaction in unrelated
languages. In particular, Jakobson claimed that mama words derived from the nasal
murmur mmm… mmm… of suckling babies; he le papa words unexplained – but
may have considered, from a far-fetched structuralist perspective, that the non-nasal
counterpart p of nasal consonant m should naturally apply to the non-breastfeeding
counterpart of the mother, namely the father.
is theory was not supported by any historical evidence. Its authors relied on the
growing body of observations of child language acquisition to build an indirect expla-
nation, along the lines of “kinship appellatives resemble each other much too much
to have arisen by chance. Since conventional wisdom has it that the many language
families they appear in are unrelated to each other, here is how they might have been
spontaneously invented in various languages, even though this process has never been
observed.” In spite of its indirectness and the good bit of wishful thinking it relied on,
this theory immediately received wide approbation and is still taught in linguistics
departments as the obvious explanation of the global distribution of papa/mama word s.
Murdock and Jakobson’s view was rst challenged 35 years later by the American
linguist Merritt Ruhlen (1994a). He had discovered a new widespread appellative kaka
‘brother, uncle’, which had escaped the attention of comparatists for a century and a
half aer the global distribution of papa/mama words had become known to linguists.
Its phonetic form was unlikely to have emerged from the babbling of babies, since
velars like k are acquired later than labials (p, b, m) and dentals (t, d, n). And it seemed
Brave new words
unlikely that its meaning had emerged independently with the same phonetic form in
many dierent language families. Ruhlen concluded that the many kaka words he had
discovered in a range of language families from Eurasia, the Americas and Oceania
had to have been inherited from a common ancestral Proto-Sapiens language. He also
suggested that there had to be an inherited component behind the global distribution
of papa/mama words as well, and that Jakobson’s explanation of their origin by conver-
gence was probably “exaggerated, if not completely mistaken” (p. 124).
Ruhlen’s discovery prompted us to undertake a global etymological comparison of
kinship appellatives. We rst checked the etymological support of Proto-Sapiens kaka,
and found literally hundreds of kaka words in parts of the world where they had not
been documented by Ruhlen, notably Africa (in the Niger-Congo, Nilo-Saharan, and
Khoisan families), Australia (in most subgroups of the Australian family), and New
Guinea (in many branches of the Indo-Pacic family), as well as in many more lan-
guage families from other continents, such as Afroasiatic, Turkic, Mongolic, Tungusic,
Uralo-Yukaghiric, Japonic, Burushaski, Sino-Tibetan, Yeniseian, Dravidian, Eskimo,
and Na-Dene, as well as probably, under a phonetically decayed form, Indo-European
(Bancel & Matthey de l’Etang 2002; see Map 1).
Map 1. e global etymology kaka ‘mother’s brother, spouse’s father, grandfather, elder
brother’ (sample data)
Languages are grouped in phyla, themselves arranged in columns according to their approximate
respective location on the planisphere. Phylum names (e.g. dene-caucasian) appear in capitals above
or below each column, followed by the most likely original form and the kinship positions it most likely
referred to; in each row, the language name (e.g. Zuñi) is followed by the vernacular word in italics
(e.g. aga), then by the abbreviated main meaning of the word (F ‘father’, M ‘mother’, B‘brother’, Z ‘sister’,
Sib ‘sibling’, S ‘son’, Sp ‘spouse’, Gd ‘grand’, Pt ‘parent’, Ch ‘child’, e‘elder’, y‘younger’, MB ‘mother’s brother’,
etc.); in Proto-Austronesian and Austronesian languages, Sib+ glosses words referring to an elder sibling
of opposite sex to ego (elder brother of a female, elder sister of a male)
Pierre J. Bancel & Alain Matthey de l’Etang
On the basis of the data from some 700 languages we had rst investigated, we also
determined that the focal etymological meaning of kaka was ‘mother’s brother’ rather
than ‘uncle’, followed by the less widespread meanings ‘grandfather’ and ‘brother’
(Matthey de l’Etang & Bancel 2002). We also suggested that kinship appellatives might
indeed be much older than Proto-Sapiens, and that their simple phonetic form and
specic use as calls by babies might have played a crucial role in the emergence of
articulate language (Bancel & Matthey de l’Etang 2002).
Further work, relying on a growing database of kinship terminologies (now com-
prising over 2,200 languages), led us to develop our theories about both the Proto-
Sapiens origin of kinship appellatives (Bancel et al. 2010; Matthey de l’Etang & Bancel
2005, 2008, in preparation; Matthey de l’Etang et al. 2010) and their role in the emer-
gence of articulate language (Bancel & Matthey de l’Etang 2005, 2008, 2010).
. Trask and the historical emergence of papa/mama words
Both Ruhlen’s and our theses, however, were soon opposed by a comparative linguist,
the late Larry Trask.1 To defend Murdock’s and Jakobson’s theory of the multiple, spon-
taneously convergent origins of papa/mama words, Tr a sk (2 00 4 ) reviewed the history
of these words in various languages, and concluded in favor of their “endless re-creation
and recycling” (p. 15). It was the rst time, in over a century and a half, that an attempt
was made to substantiate the traditional theory from a historical viewpoint. Indeed,
Trask’s work was useful in forcing us to descend from a global, essentially statistical,
viewpoint to the level of individual languages and families, in order to show that these
words, contrary to Trask’s claims, are not innovations in any particular language, but
have been preserved throughout the histories of their respective families. Expanding on
a previous answer (Matthey de l’Etang & Bancel 2008), our main task hereaer will be
to show, with a wealth of comparative data, that Trask’s study is awed by fundamental
fallacies, that none of his examples is an innovation, and that all of them are, instead,
words that have been preserved over millennia with little or no change.
.. Inherited papa/mama words in Indo-European languages
By a radical misinterpretation, Trask confuses papa/mama with father/mother words.
All his examples of lost or decayed papa/mama words are in fact father/mother words,
. Trask did not quote our work or Ruhlen’s, but there is little doubt that his study was in-
tended as an answer to it, as it was published two years aer our first papers had appeared
(Bancel & Matthey de l’Etang 2002; Matthey de l’Etang & Bancel 2002) in the comparative
linguistics journal Mother Tongue, of which Trask was an assiduous reader and contributor,
always to defend the traditional view that no trace of common linguistic inheritance older
than a few millennia should be taken seriously.
Brave new words
normal words of the standard adult lexicon, used to refer to any parent rather than to
address one’s own. Let us begin with the Indo-European family,2 from which he draws
numerous examples of “innovated” papa/mama words.
e Proto-Indo-European (PIE) words *patēr ‘father’ and *matēr ‘mother’ are, as
Trask (2004, p. 12) himself says,
mama/papa words which have acquired a sux -ter…. Already these words were
being treated like other words in the language. Since PIE, the original words for
‘mother’ and ‘father’, where they have survived at all, have undergone the usual
changes in pronunciation in the languages possessing them,
like Swedish far and mor, French père and mère, or Irish athair (phonetically [ahir])
and mathair [ma:hir]. So Trask concludes:
It is scarcely likely that anyone would recognize [ahir] as a mama/papa word,
but in origin it denitely is. e mama/papa words are in no way resistant to the
process of linguistic change, including regular changes in pronunciation. Nor are
they resistant to loss. (Trask 2004, p. 12)
ere is not the least doubt that PIE *patēr and *matēr, evidently derived from preex-
isting *pa(pa) and *ma(ma), “in origin denitely” were papa/mama words.
But in origin only. Already in PIE, *patēr and *matēr were no longer papa/mama
words – simple reduplicative words mimicking the babbling of babies and used to
address one’s own parents. Instead, they had become father/mother words – ordinary
words of the PIE lexicon, used to refer to anyone’s parents, as are all their derivatives
in modern languages: English father and mother, German Vater and Mutter, Swed-
ish far and mor, Icelandic faðir and móðir, French père and mère, Spanish and Italian
padre and madre, Occitan paire and maire, Irish athir and mathir, Greek patéras and
mitéra, Armenian hayr and mayr, Persian padar and mādar, Ossetic fyd and mad, and
hundreds of others. Word replacement and phonetic change had to – and obviously
did – apply normally to these normal words of the adult lexicon.
And *patēr and *matēr certainly were not the rst words of PIE-speaking children
some 7,000 years ago, any more than father and mother are the rst words of English
children today, or père and mère those of French children.
. e Indo-European language family, whose discovery in the end of the 18th century and
further exploration in the 19th gave birth to linguistic science, comprises most groups of
languages spoken in Europe today (Celtic, Italic, Germanic, Baltic, Slavic, Albanian, Hellenic,
Armenian), as well as the huge Indo-Iranian group, itself divided in three subgroups (Indic,
Nuristani, and Iranian); it also includes two extinct groups, Anatolian and Tocharian. e
reader unfamiliar with language classification will find members of each group listed in Ap-
pendices A to C and G, with examples of common words.
Pierre J. Bancel & Alain Matthey de l’Etang
Moreover, Trask does not document a single papa/mama word known to be lack-
ing in a given stage of a language’s history, which appeared in a subsequent stage. He
merely assumes that, in every language where papa/mama and father/mother words
coexist, the former must be more recent than the latter. And in doing so, he oen goes
against their known etymology.
As we will see in detail below, all of his “new” papa/mama words have been inher-
ited from the most ancient stages of their respective language families. When PIE
*patēr and *matēr were derived as reference terms from Pre-PIE *pa(pa) and *ma(ma),
the more ancient forms did not disappear. *Pa(pa) and *ma(ma) must have been kept
in parallel use as appellatives, just as in English father coexists with dad, and mother
with mom. e reader is referred to Appendix A, which displays the etymological
series supporting PIE *ma(ma) ‘mother, mom’ in the Tower of Babel3 Indo-European
database (Nikolayev 2007), completed by other data. From Prakrit māmikā ‘mother’,
Classical Greek mā! (gā!) ‘(Earth) Mother’ and Latin mamma ‘mommy’ to Punjabi mā"
~ mā"u ~ māī ~ māmmī ‘mother’, Persian mām ‘mom’, Armenian mam ‘grandmother’,
Modern Greek mama ‘mom’, Ukrainian mama ‘mom’, Latvian mãma ‘mom’, Faeroese
mamma ‘mom’, Sutsilvan Rumantsch moma ‘mom’, French maman ‘mom’, mamie ~
mémé ‘granny’, Breton mam ‘mother’, or Gheg Albanian mam# ‘mother’, more than a
hundred languages from the vast majority of IE subgroups unambiguously establish
the PIE antiquity of this word.
Nikolayev (2007) does not posit a PIE root *pa or *papa. Its existence, however,
cannot be doubted given the comparative data of Appendix B, which provides some
170 papa words from well over a hundred IE languages, from Palaic pāpa ‘father’,
Prakrit bappa ‘father’, Khwarezmian papa ‘father’, Classical Greek pappa ‘dad’, pappous
‘grandfather’ or Latin pappa ‘dad’ and pappus ‘grandfather’, to Marathi bāp ‘father’,
Kâmv’iri vov ‘grandfather’, Farsi bâbâ ‘father, grandfather’, Armenian pap ‘granddad’,
Modern Pontic Greek papa ‘dad’, Latvian paps ‘dad’, Danish papa ‘dad’, or Occitan papà
‘dad’ and papet ‘grandfather, granddad’.
e Proto-Indo-European descent of these words eliminates many of Trask’s
“innovations”: Greek mama, Icelandic mamma and pabbi, Norwegian mamma and
pappa, French maman and papa, Italian mamma and babbo, Polish mama, Bengali ma
and baba, Hindi baba or bap, Persian mām and baba, Latvian mama and paps (Trask
. e Tower of Babel Project 〈http://starling.rinet.ru/〉 brings together the Russian State
University of the Humanities (Moscow, Russia), the Moscow Jewish University (Russia), the
Russian Academy of Sciences, the Santa Fe Institute (New Mexico), the City University of
Hong Kong (China), and the Leiden University (e Netherlands). It provides free access to
etymological databases for numerous language families, compiled by some of the best special-
ists worldwide. In our etymological lists, unreferenced data not drawn from Nikolayev (2007)
may be found in easily accessible standard dictionaries.
Brave new words
2004, pp. 13–14) all directly derive from PIE appellatives *mama and *papa, which
themselves must be even older than PIE, since in PIE times their derivatives *patēr
‘father’ and *matēr ‘mother’ were already well established.
Two double examples, jointly presented by Trask to illustrate the converging pro-
cess of innovation in kinship appellatives, are worth special consideration:
e ancestral PIE words [*matēr and *patēr] have been completely lost in a
number of the daughter languages, lost and replaced by other words. Two of
those languages are Romanian and Welsh […]:
‘mother’ ‘father’
Romanian mama tata
Wel sh mam tad
But look at the words which have replaced the lost older ones! e newer words
which have replaced the older ones are themselves mama/papa words. According
to the Proto-World account, … [t]he mama/papa words are supposed to be no
more than ancient survivals, and they can’t do anything except survive for a while
longer or disappear. ey absolutely can’t reappear in languages which have lost
them. But they do. And they do it all the time. (Trask 2004, p. 12)
Reappear all the time? Romanian mámă certainly did not (re)appear out of the blue,
nor did Welsh mam. e data in Appendix A establish that they were inherited from
Latin mamma and Proto-Celtic *mama, respectively, and that both ultimately derive
from PIE *mama. ey are exactly the “ancient survivals” Trask does not want to see in
them. And this has long been known to etymologists (Romanian: Meyer-Lübke 1911;
Academia Română 1998; Welsh: Charles-Edwards 1993, p. 169).
But what about Romanian tátă ‘father, dad’, and Welsh tad ‘father’? Could they
be “newer words which have replaced the older ones”? ey could not. Romanian
tátă has been known for a century to derive from Latin tăta ‘dad’ (Meyer-Lübke 1911;
Ciorănescu 1958–66; Academia Română 1998). According to Charles-Edwards (1993,
p. 169), Welsh tad goes “back at least to the Romano-British period” (43CE to early
5thcentury), as it is found in all the ancient stages of the Brythonic group of Celtic
(Old Cornish, Middle Welsh, and Middle Breton). And Old Irish (a language belong-
ing to the Goidelic group) data ‘foster father’ shows that the word must be of Proto-
Celtic origin.
But their antiquity in their respective language groups – Romance and Celtic – is
not the end of their story. Both words belong to the PIE etymology *tata ‘dad, father’
reported in Appendix C, again based on Nikolayev (2007) and completed with data
from various sources. Once more, from Hieroglyphic Luwian tati(a)- ‘father’, Vedic
Sanskrit tatá ‘father’, Old Avestan tā ‘father’, Classical Greek tatā ‘daddy’, or Latin tăta
‘dad’ to Kâmv’iri tot ‘father’, Roshani taat ‘father’, Czech táta ‘father, dad’, Latvian tēte
‘dad’, Romanian tátă ‘father, dad’, Breton tad ‘father’, or Albanian tat# ‘father’, both
Pierre J. Bancel & Alain Matthey de l’Etang
ancient and modern data from most subgroups abundantly testify to its inheritance
from the earliest PIE stages.
To s u m u p, Tr a s k’s c l a im t ha t R o ma n i a n tátă and Welsh tad are words that have
recently (re)appeared is, again, contrary to obvious etymological and comparative facts.
.. Inherited papa/mama words in Dravidian and Turkic languages
Let us also consider two non-Indo-European examples cited by Trask, in Tamil and
Turkish, respectively. In Trask’s (2004, p. 14) view, the “informal” Tamil word appaa
‘dad’ is newer than the “formal” takappan$ ‘father’. But it simply cannot be. e hon-
oric takappan$ is a compound formed from tak, an adjective form of verb taku ‘to be
excellent’, and appan$ ‘father’ (Emeneau 1953, p. 342, 10). And Tamil appan$ ‘father’ is
itself a suxed derivative of appaa ‘dad’, just as PIE *patēr was a suxed form of *pa-,
as revealed by the comparative data in Appendix D, drawn from the classical etymo-
logical dictionary of Dravidian.
e earliest trace of Tamil appaa is found in a 3rd-century CE inscription, used as
a masculine honoric sux (Mahadevan 2003, p. 609), as in Modern Kannada, Tulu,
and Telugu. And appaa evidently derives from Proto-Dravidian.
With regard to Turkic, as Trask himself says, the inherited word for ‘father’ is ata,
and this word is “still the everyday word in most Turkic languages.” But, in Modern
Turkish,
the word ata [has become] an elevated word meaning ‘forefather, ancestor’, [and]
the everyday word for ‘father’ is now baba. is, of course, is another mama/papa
word, and it used to be the Turkish word for ‘daddy’, but now it is the ordinary
word for ‘father’, and ‘daddy’ must now be expressed by adding a diminutive
sux, producing babacik. (Trask 2004, p. 13)
To Trask, this succession of a meaning shi, a replacement, and a suxation illustrates
the idea that nursery words change ceaselessly. Proto-Turkic *ata ‘father’ is indeed
reected in many ancient and modern Turkic languages, from Old Uighur ata to Sary-
Yughur ata through Tuvin a’da, Azeri ata, and Khakassian ada, all meaning ‘father’
(Appendix E1). Aer 1,300 years, most of these terms remain identical to Proto-
Turkic. In Turkic languages, preservation of *ata has been the rule. Furthermore, its
meaning shi to ‘forefather, ancestor’ in Modern Turkish is quite a common one. One’s
father is one’s closest male ancestor, and in nearly all languages words meaning ‘father’
may also refer to other male ascendants, or even brothers and male descendants. is
was the source of a vast majority of their semantic changes, which are merely expan-
sions or retractions of their scope within the narrow eld of kinship relationships,
mostly within the same gender.
For its part, the diminutive babacığım ‘daddy’ (rather than babacik, which is not
found in a single Turkish dictionary) does not replace baba (found in all Turkish
Brave new words
dictionaries with the meaning ‘dad’) in Turkish children’s rst words nor in their
parents’ baby talk, any more than English daddy replaces dad, Italian babbino replaces
babbo, or French papoune or papounet replaces papa. ey are aectionate diminu-
tives, and may continue to coexist for centuries with their respective root words baba,
dad, babbo, and papa, or perhaps enter the standard language with a new meaning. But
baba, dad, babbo, and papa will remain, because babies need them to learn to speak,
and parents to teach children, as will be explained in Section 3 below.
Finally, as for Turkish baba itself, far from being new, it was borrowed from
Persian (Nişanyan 2001) aer the Türks invaded the Persian Empire, a borrowing cer-
tainly facilitated by the existence in Turkish of another old Turkic word, aba ‘father,
ancestor’, preserved in many Turkic languages (Appendix E2).
Borrowed? Yes, papa/mama words may be borrowed, and indeed they are prob-
ably more frequently borrowed than any other words in the basic lexicon. We have
already met Greek baba, borrowed from Turkish – which had previously borrowed
it from Persian. Albanian baba was also borrowed from Turkish during the Ottoman
domination over the Balkans. English dad, an isolated form in the Germanic group of
Indo-European, whose other members all have papa forms (Appendix B), was likely
borrowed from Brythonic Celtic, where tad ~ tat forms are general (Appendix C),
when the Anglo-Saxons invaded Great Britain. It is also likely that Romanian tátă
‘father, dad’, a descendant of Latin tăta ‘dad’, which even replaced in Romanian the
outcome of Latin pătĕr ‘father’, was helped to survive – and thrive – by the forms tata
‘father, dad’, which are general in the surrounding Slavic languages from which Roma-
nian borrowed thousands of other words, while many other Romance languages lost
Latin tăta and preserved pappa instead.
But borrowing is not an innovation, in the sense of a newly created word.
A borrowed word has a history in the donor language, and the receiver language con-
tinues this history. In the case of Turkish baba, as we have seen, its Persian source
derives from Proto-Indo-Iranian *baba, itself from Pre-Proto-Indo-European *papa
( Appendix B). Brave new word.
.. Inherited papa/mama words in Chinese languages
e case of Chinese, not studied by Trask, also deserves consideration. In nearly all
modern Chinese languages, from Mandarin to Cantonese, address terms used for one’s
father and mother are pa ‘dad’ and ma ‘mom’, respectively (see Appendices F1 and F3).
Only their tonal contours vary according to dialect. Both pa and ma have reduplicated
variants, respectively papa and mama, felt to be more childish by speakers (Agnès
Gaudu, personal communication).
In the Chinese Characters (Starostin 2006) and Modern Chinese Dialects (Wang
2004) Tower of Babel databases, modern pa ‘dad’ forms are assigned an etymology
dating back to Preclassic Old Chinese paʔ ‘father’, implying very little variation over
Pierre J. Bancel & Alain Matthey de l’Etang
some 3,500 years (Appendix F1). e ancient forms are apparently shared with the ety-
mology of the referential word ‘father’ (Appendix F2), derived from Middle Chinese
pǘ, a dialectal form attested since the Tang period (seventh to tenth centuries CE),
evolved into Beijing, Jinan, or Xi’an fu3, Shuangfeng 'əu32, Chaozhou pe22, Fuzhou
xu32, Shanghai vu32, etc. (superscript numbers indicate tones).
In the phonetic form paʔ ‘father’ of Preclassic Chinese reconstructed by Starostin
(2006), the nal glottal stop -ʔ is essentially posited to explain the tonal evolutions in
modern dialects, inspired from regular correspondences in other words. But in words
like ‘dad’ and ‘mom’, the evolution of whose tonal pattern is highly likely to have been
inuenced by expressive intonational patterns, this particular nal -ʔ does not need
to ever have existed. And it surely did not, given that it is not present in even a single
modern Chinese dialect.
Indeed, what happened in Chinese seems clear. From a Preclassic pa, a form
pwá ~ pwó ‘father, dad’ appeared during the Han period and progressively special-
ized as a referential term, giving rise to Middle Chinese pǘ ‘father’, from which
all the modern forms fu ‘father’ derive. Meanwhile, pa continued to be used as an
address term in the spoken language and was transmitted without any change in all
Chinese dialects. However, the pictogram that originally read pa received the pho-
netic reading of the reference term, showing that pa was originally used for both
address and reference.
e identical (except for tones) pa forms of all modern dialects – known not only
by ideograms, but by phonetic descriptions as well – prove that pa survived unchanged
throughout the history of Chinese. And the Eastern Han and Postclassic pwá or pwó
forms have been misattributed in Starostin’s database – they may not be the phonetic
ancestors of modern pa ‘dad’ forms but are forerunners in the evolution of modern fu
‘father’ forms.
A similar situation appears in the etymology of terms meaning ‘mom’ and
‘mother’, although the two terms may have begun to dierentiate already in Preclas-
sic Chinese (see Appendices F3 and F4). In ancient forms, again, the aspirated initial
mh- and the nal glottal stop -ʔ are reconstructed on the basis of tonal developments in
modern dialects. us, as in the case of pa ‘dad’, both are far from assured and indeed
are superuous, given their absence from ma words in all modern dialects and the
expressive uses of appellatives.
Just as Proto-Indo-European speakers created *patēr ‘father’ and *matēr ‘mother’
from preexisting *pa(pa) and *ma(ma) words and continued to use them in paral-
lel, Chinese speakers developed new reference (father/mother) words from the Old
Chinese words pa and ma that were initially used for both reference and address. But
speakers continued to use in parallel the original pa/ma forms as address terms. e
new reference terms have strongly evolved in modern Chinese dialects, e.g. Wenzhou
voy22 or Shuangfeng 'əu32 ‘father’, or Chaozhou bo21 ‘mother’ – in which a non-nasal
Brave new words
consonant has even appeared – but in all dialects the address terms pa ‘dad’ and ma
‘mom’ have remained exactly the same as those used over 3,000 years ago.
. Summary
– Papa/mama words are exempt from most phonetic evolutions, but may, on occa-
sion, vary phonetically within the limits allowed by babbling as regards vowel
quality and length as well as consonant gemination and the voiced/voiceless (or
fortis/lenis) contrast, for example within the Germanic group, German Papa,
Rhine Franconian Pàppe ~ Bàbbe, Bavarian Babba, Faeroese pápi, Icelandic pabbi
‘dad’.
– Due to their once common use to address elders respectfully or youngsters aec-
tionately, they may vary semantically, in general within the same gender, e.g.
Sogdian bâbay ‘father’, Yaghnobi (a modern descendant of Sogdian) bobo ‘grand-
father’, but may also occasionally be recruited into morphological alternations,
e.g. Bashkarik mêm ‘mother’s mother’, mâm ‘mother’s father’, locally introducing
etymological confusion.
– ey may give rise to father/mother words and continue to coexist with them,
for example, Pre-PIE *papa/*mama having given rise to PIE *papa/*mama and
*patēr/*matēr, or Old Chinese pa/ma having evolved into Mandarin pa3/ma11 and
fu3/mu2.
– Papa/mama words may be borrowed, such as Modern Greek baba ‘dad’, borrowed
from Turkish. Such borrowings are most of the time facilitated by similar preex-
isting words in the target language: Homeric pappa > Hellenistic papa, preserved
in Modern Pontic, diered from Turkish baba only in consonant voicing. In turn,
Old Turkish aba ‘father’ diered from Persian baba, borrowed into Turkish, only
in partial versus full reduplication.
– Ancient languages possessed more papa/mama words than modern ones and
used them extensively as terms of respect for elders. Certainly, PIE *papa and
*tata were not exact synonyms; otherwise they would not have been preserved
in so many of the descendant languages, always with dierent meanings in lan-
guages that preserve both. eir original semantic dierence, which may have
resided in their connotations rather than the persons they referred to, remains
uncertain. Due to semantic overlap and the loss of importance of kinship rela-
tions – which used to be the very essence of the social organization in all hunter-
gatherer societies, the only way of life of all human beings until some 10,000 years
ago – some of these words have been lost in historical times, as was Latin tăta ‘dad’,
lost in French, Occitan, Spanish, and Portuguese.
– A few such words, however, do randomly appear in the course of the history of
individual languages, such as French tata ‘auntie’, a diminutive of tante ‘aunt’. Such
Pierre J. Bancel & Alain Matthey de l’Etang
cases do not stem from babies’ babbling but from adults’ baby talk. Very impor-
tantly, they do not obey the distribution rule Oral stops for males, nasal stops for
females. is rule was already observed in 75% of languages by Murdock (1959)
in his survey of words meaning ‘father’ and ‘mother’ in 474 languages, and was
conrmed by our own statistics bearing on 1,184 languages (Table 1; for a detailed
analysis, see Bancel et al. 2010). And, given the massive preservation of original
forms in most languages from all families, innovations may only represent a tiny
minority of the countless papa/mama words worldwide.
Table 1. Most prominent meanings of papa, kaka, nana, and mama
Form
Meaning
PAPA KAKA NANA MAMA
Number
of occ
Percent
(tot. >
100)
Number
of occ
Percent
(tot. >
100)
Number
of occ
Percent
(tot. >
100)
Number
of occ
Percent
(tot. >
100)
F 288 } 59.2% 12 } 2.4% 38 } 6.2% 84 } 26.1%
F + FB 106 3 4 82
FB 100 15.0% 59 9.2% 1 0.1% 7 1.1%
FZ 36 5.4% 10 1.6% 48 7.1% 31 4.9%
M20 } 6.0% 14 } 3.0% 250 } 64.0% 232 } 43.1%
M + MZ 20 5 182 42
MZ 8 1.2% 24 3.7% 48 7.1% 49 7.7%
MB 33 5.0% 221 34.5% 11 1.6% 105 16.5%
B+ 100 15.0% 111 17.3% 28 4.1% 4 0.6%
Z+ 27 4.1% 64 10.0% 59 8.7% 9 1.4%
Sib+ 7 1.1% 32 5.0% 17 2.5% 9 1.4%
GdF 134 20.1% 86 13.4% 16 2.4% 15 2.4%
GdM 45 6.8% 61 9.5% 48 7.1% 52 8.2%
GdPt 15 2.3% 35 5.5% 4 0.6% 12 1.9%
GdPt +
GdCh
42 6.3% 31 4.8% 4 0.6% 35 5.5%
GdCh 38 5.7% 28 4.4% 0 0.0% 6 0.9%
Ch 14 2.1% 3 0.5% 65 9.6% 35 5.5%
total out
of 1,184
languages
1,033 cognates
in 666 languages
(56% of sample)
799 cognates
in 641 languages
(54% of sample)
823 cognates
in 675 languages
(57% of sample)
809 cognates
in 635 languages
(54% of sample)
Figures calculated for 1,184 languages; percentages have been calculated with regard to the number of
languages attesting one or more words in the series concerned. Not all kinship relations attested for each
term are listed above: for each series, at least a dozen other relations are attested by a few items. As of
August 2010, our database comprised more than 2,200 kinship terminologies, and percentages would not
be very dierent. (Table from Bancel et al. 2010)
Brave new words
Both Murdock’s semantic convergence rates and our own statistics have been calcu-
lated for father/mother words and papa/mama words taken together. e reason is, as
we found in our own compilation of kinship terminologies, that while words meaning
‘father’ and ‘mother’ nearly always gure even in the shortest wordlists noted by eld
linguists or anthropologists, the corresponding appellatives are seldom noted – prob-
ably because of their perceived childish nature and their near identity in all languages,
which result in a kind of disdain towards them. Since father/mother words are much
less stable than papa/mama words, there is no doubt that the proportion of papa/
mama appellative words complying with the oral/nasal distribution rule would be
much higher than 75%, and probably above 90%.
However, before concluding that papa/mama words must share a common origin,
we have to address another possibility.
. Chance resemblances?
e main argument opposed to etymologies linking languages at a greater remove
than Indo-European or other relatively recent ancestor languages is that the compara-
tive series they rely on might have arisen by chance. To the best of our knowledge,
this argument has never been leveled at papa/mama words, and we might consider
it discarded in advance by the wide acceptance of Murdock’s and Jakobson’s theory
of their spontaneous convergence under the inuence of babies’ babbling. If chance
might have led to this convergence, putting forward or accepting any other explana-
tion would have been absurd.
Indeed, the true absurdity would be to consider that the massive global conver-
gence of papa/mama words could have arisen by chance. e overwhelming majority
of these words are traceable to the very origin of their respective language family, in
which they have survived for millennia – in the case of Indo-European languages,
for 6,000 to 8,000 years, according to the most likely estimations. How could they
have spontaneously emerged in dierent families all over the world with convergent
meanings and phonetic forms in a distant past, while they have been among the most
conservative words in the last several millennia? In ancient languages as well, they had
to have been inherited, even if some may have been borrowed in a minority of cases.
e primordial role of kinship in the social organization of all peoples before
the appearance of agriculture – and undoubtedly for eons, as clear precursors of kin-
ship relationships are found in our closest ape cousins, chimpanzees and bonobos (de
Waal 1982; Fouts 1997) – excludes the possibility that they could be recent inventions.
eir global distribution denitely excludes generalized, intercontinental borrowings,
so that the only remaining explanation is that they have been transmitted over doz-
ens of millennia from a common ancestor language. ey may only have stemmed
from a common, Proto-Sapiens origin – an idea which makes sense with regard to
Pierre J. Bancel & Alain Matthey de l’Etang
both archeological and genetic data about the expansion of Homo sapiens from their
African homeland some 50,000 to 100,000 years ago.
We could thus dispense with a detailed refutation of the chance hypothesis. Nev-
ertheless, we will address it here in some detail, as we are convinced that deep-time
linguistic comparison has much more to tell us about the development of human lan-
guage as far back as the beginning of our species’ expansion, thus shedding light on
a crucial period in Homo sapiens’ history, with the dramatic acceleration of technical
evolution and the appearance of food cooking and of personal and graphic ornaments.
Many scholars think that these changes must be linked to an evolution of human lan-
guage ability, with the most frequently mentioned candidate being the emergence of
syntactic articulation. We have no doubt that the comparative-historical study of lan-
guages can help to understand this evolution, and we will illustrate this opinion at the
end of this article (Section 5.1).
rough a detailed analysis of two tentative probabilistic refutations of deep-time
etymologies, we will show that proving or disproving Proto-Sapiens etymological
series by means of probabilities would demand calculations involving many param-
eters, some of which are not easily amenable, if at all, to numerical representation.
It will also appear that the etymologies subjected to these treatments are beyond the
point where a probabilistic assessment is necessary. Similarly, regular phonetic corre-
spondences in low-level linguistic families are far beyond the level where chance might
be involved and are with good reason regarded as indisputable proof of the common
descent of the words they are found in, without having ever undergone any kind of
mathematical assessment.
. Inaccurate calculations
e probabilistic refutations of deep-time linguistic comparisons known to us fall into
two categories. e rst one is that of historical linguists unfamiliar with the basic
principles of probabilities. For instance, the Indo-Europeanist Donald Ringe (2002),
trying to show that Greenberg’s (2000) Eurasiatic4 etymologies are due to chance
resemblances, overlooks the fact that a probability is a ratio – that is, it describes the
number of chances for a particular event to happen out of a total number of possible
events, so that one has 1 chance out of 6 of getting an ace when throwing an ordinary
die, but only 4 out of 52, or 1/13, when taking a card from a deck. is leads Ringe, in
six dense pages, to multiply probabilities as he adds parameters that obviously shrink
them – as if he had found that, when taking a card from each of four decks, there were
. Eurasiatic is a macrofamily of languages discovered by Greenberg (2000–2001) encom-
passing the Indo-European, Uralo-Yukaghir, Altaic, Koreo-Nippo-Ainu, Gilyak, Chukchi-
Kamchadal, and Eskimo-Aleut language families.
Brave new words
4 × 4 = 16 “absolute” chances of getting 4 aces, instead of (1/13)4 = 14/134 = 1/28,651,
or 1 chance out of nearly 30,000. As a result, Ringe nds that Greenberg had “more
than 35 quintillion” chances of discovering a rst-person pronoun root *m- common
to 21 language groups from northern Eurasia.5 Out of how many possible outcomes,
he does not mention, not realizing that 35 quintillion chances out of 3,500 quintillion
would yield a tiny probability of 1%, or 0.01, while if there were 35 octillion possible
outcomes, it would descend to a minuscule probability of 1billionth, or 1/109. No reli-
able conclusions can be drawn from such fanciful calculations.
. Inaccurate comparative linguistics
e second category of erratic probabilities is due to scholars unfamiliar with
comparative-historical linguistics performing apparently correct probabilistic calcula-
tions on irrelevant parameters. is is what the phonetician Louis-Jean Boë does with
Bengtson and Ruhlen’s (1994) global – that is, Proto-Sapiens – etymologies, in a study
whose successive versions (Boë et al. 2003; Boë 2004; Boë et al. 2006) do not show any
real improvement in this regard.
.. Inaccuracy with regard to linguistic taxonomy
Knowing the proportion of languages that reect an assumed original root seems
important to ensure that the assumed cognate words are not random look-alikes: if
you take a card from each of 52 decks, you may be nearly sure of getting at least one
ace, and the greatest probability is that you will get four of them. How does this work
with languages? Boë et al. (2003) count the total number of languages mentioned by
Bengtson and Ruhlen (1994) in support of all their 27 Proto-Sapiens etymologies.
ey nd 1,317 of them, and, assuming that this was the total number of languages
investigated by Bengtson and Ruhlen, they relate to this total the average number of
languages cited in support of each etymology. ey nd that each etymological series
comprises an insucient number of languages and families. But their count and its
. is first-person root m- is represented in English by me, my, mine, and as a relic of the
PIE conjugation system in I am. In an unpublished study, we have found that it survived as
the first-person pronoun root in 99.6% of 494 Indo-European languages and dialects, from
Icelandic mig to Assamese mōk through Portuguese me, Greek mè, Russian menja or Pashto
mā, whose common descent from the PIE root *m- is acknowledged by all Indo-Europeanists,
including Ringe himself. Only two IE languages, Tocharian A and B, may have lost it. is
stunning preservation, paralleled in most of the 20 other families alluded to by Ringe, from
Turkic to Eskimo through Finno-Ugrian and Chukchi-Koryak, shows that chance has nothing
to do with the presence of this pronominal root in 21 families, most of which also share a
second-person root t- (English thou, thee, thy, thine) as well as some 70 other grammatical
roots and hundreds of lexical roots (Greenberg 2000–2001).
Pierre J. Bancel & Alain Matthey de l’Etang
alleged consequences are simply pointless. When introducing their Proto-Sapiens ety-
mologies, Bengtson and Ruhlen warn that the potential descendant words they quote
are only examples:
[S]ince the existence of these roots as characteristic features of the language
families cited has already been established by other scholars, and is not for the
most part in question, we do not give the complete documentation for each family,
limiting ourselves in most instances to an indication of the range of semantic
and phonological variation within the family. e reader who wishes to see every
relevant form for a given family should consult the sources cited.
(Bengtson & Ruhlen 1994b, p. 291; emphasis added)
Let us illustrate Boë et al.’s misinterpretation of Bengtson and Ruhlen’s data – not for
each of the 27 global etymologies, because that would take several books, nor even for
a single one, but for a single family supporting a single etymology. In support of their
Proto-Sapiens etymology tik ‘nger, one’, Bengtson and Ruhlen (1994, pp 322–323)
give 184 reexes from 165 languages (12.5% of the 1,317 languages they quote), includ-
ing a mere 9 reexes taken from only 6 Indo-European languages:
Indo-European: Proto-Indo-European *deik- ‘to show, to point’, *dekm)- ‘ten’;
Italic: Latin dig(-itus) ‘nger’, dic(-āre) ‘to say’, *decem ‘ten’; Germanic: Proto-
Germanic *taihwō ‘toe’; Old English tahe ‘toe’; English toe; Old High German
zêha ‘toe, nger’. (Bengtson & Ruhlen 1994, p. 322)
Does this sample exhaust what Bengtson and Ruhlen could have found in the Indo-
European family? Well, not exactly. Appendix G1 displays the data mentioned by
Nikolayev (2007) under the PIE etymology *deik’e- ‘to show, to point’, completed by
Pokorný (1959), Lubotsky (no date), Turner (1962–1966) and standard dictionaries of
various modern languages. While it is still far from exhaustive, it oers 170 derivatives
of the Indo-European root *deik’e- ‘to point, to show’ in some 80 languages. As regards
PIE *dekm) - ‘ten’, Appendix G2 lists 250 reexes from 247 languages, drawn from the
same sources plus the remarkable compilation of Rosenfelder (no date). e common
descent of these words is assured by two centuries of Indo-Europeanist comparison
and, as Bengtson and Ruhlen say, is “not for the most part in question.”6
us, in the Indo-European family alone, over 400 possible reexes of Proto-
Sapiens tik add to the 9 examples given by Bengtson and Ruhlen. And Indo-European
is but one of the 21 families displaying reexes of Proto-Sapiens tik ‘nger, one’ in
Bengtson and Ruhlen’s series.
. Only Classical Greek dak-tulos ‘finger’ and its direct Modern Greek descendant ðak-tilo
‘finger’ are not recognized by Indo-Europeanists as related to the series (e.g. Chantraine 1968,
pp. 249–250) because of their irregularity; we nevertheless think they do belong to it.
Brave new words
Boë et al.’s claim, based on language counts, that Bengtson and Ruhlen’s etymolo-
gies are insuciently supported, and thus likely to have resulted from chance resem-
blances, obviously falls far o the mark.
Now, should Bengtson and Ruhlen have published such huge lists for all families
supporting each of their etymologies? From the viewpoint of reconstruction, no. e
two PIE roots *deik’e- ‘to point’ and *dekm) - ‘ten’ rely on regular phonetic correspon-
dences attested in innumerable other etymological series; hence their validity does
not depend primarily on the number of reexes but on the regularity in the detail
of correspondences. No lists such as those in Appendix G have ever been published
by any Indo-Europeanist, and this essentially underscores the vacuity of probabilistic
calculations that do not take into account the fact that Proto-Indo-European is an
ancestor language. With regard to the earlier history of a particular word, PIE repre-
sents all its descendant languages – those that preserved the word in question as well
as those that lost it. If a word existed in PIE, the fact that it disappeared from 4, 40, or
400 descendant languages is irrelevant to the ancestry of this word before PIE, and Boë
et al.’s method, beyond their misreading of Bengtson and Ruhlen’s warning about the
incompleteness of their examples, entirely misses this crucial point. Yet Bengtson and
Ruhlen are quite explicit once again:
A common criticism is that, with around 5,000 languages to choose from, it
cannot be too hard to nd a word in some African language that is semantically
and phonologically similar to, or even identical with, some word in an American
Indian language. … But this sort of mindless search is exactly the reverse of how
the comparative method proceeds. e units we are comparing are language
families, not individual languages. … So instead of drawing our etymologies from
thousands of languages, we are, rather, limited to [32] families, some of which
have no more than a few hundred identiable cognates. e pool of possibilities
is thus greatly reduced, and accidental look-alikes will be few.
(Bengtson & Ruhlen 1994, pp. 279–281; emphasis in the original)
e inequality of languages and proto-languages with regard to their early history also
aects contemporaneous languages: for instance, a reex found in a language such as
Basque or Burushaski, which by themselves constitute long-isolated language families,
cannot be given the same etymological weight as a reex found in one of the sev-
eral hundred Romance or Germanic dialects. is evolutionary hierarchy is not easily
reduced to gures – in particular with regard to disputed taxa, as is oen the case of
subgroupings within accepted families, and nearly always for remote macrofamilies
and phyla: should Basque be given the weight of a completely isolated language, as if
the Basques had independently discovered articulate language, or should it be consid-
ered a member of the Vasco-Caucasian macrofamily, or of Dené-Caucasian, a hotly
disputed phylum whose huge range spans across northern Eurasia far into northwest-
ern North America?
Pierre J. Bancel & Alain Matthey de l’Etang
Still, from a probabilistic viewpoint, the number of languages in which a word
from a proto-language did survive may not be entirely irrelevant to its earlier antiquity.
e two lists in Appendix G tell us that the two PIE roots *deik’e- ‘to point’ and *dekm) -
‘ten’ are among the words that have best resisted loss in the history of IE languages. In
itself, this resistance shows that these words are able to survive over long periods of
time, which is a strong a priori argument in favor of their ability to have survived over
the times that preceded PIE as well. For this reason, Bengtson and Ruhlen might have
published the detailed support of at least one of their etymologies.
But, whatever the amount of sources and data, we do not see how the taxonomic
ranking of languages (i.e. the inequality between an ancestor language and its descen-
dants, or between a long-isolated language and a dialect in a large family) could be
taken into account in a statistical calculation. e recent achievements of cladistics,
involving sophisticated probabilities, tend to show that it might perhaps be possible;
but it would demand a serious collaboration between qualied statisticians and com-
parative linguists.
.. Inaccuracy with regard to phonetic correspondences
Boë et al.’s probabilistic assessment of the phonetic validity of Bengtson and Ruhlen’s
series is inaccurate as well. ey total the dierent phonetic forms assumed by Bengtson
and Ruhlen to descend from each root (Boë et al. 2003, p. 2707), and nd it so large
that, in their opinion, any correspondence would be allowed, and thus meaningless.
e case of Proto-Sapiens tik ‘nger’, raised by Boë (2004) to illustrate Bengtson and
Ruhlen’s phonetic laxity, is again enlightening (Table 2).
According to Boë et al. the large number of dierent sounds reecting each origi-
nal sound (20 for t, 21 for i including diphthongs and loss, or even 26 if long vow-
els are counted separately, and 23 for k including loss) reveals Bengtson and Ruhlen’s
laxity in selecting their reexes. And this laxity, of course, has severe probabilistic
consequences.
But a glance at the phonetic nature of the sounds reecting each sound in t-i-k
shows that they form consistent sets, each dened by the region of the mouth where
its member sounds are formed. Since a great majority of consonant evolutions pre-
serve the original place of articulation, these sets thus encompass sounds most likely
to evolve into one another.
All consonants reecting the initial coronal consonant t- of tik are also coronals.
Coronals constitute a class of sounds pronounced with the tip of the tongue raised
close to or against the upper front teeth (interdentals, dentals) or just behind them
(alveolars, post-alveolars). ese consonants articulated in the same region of the
mouth as t are known to derive from earlier t’s in numerous languages. Not a single
labial such as p, b, p’, β, f, or v, nor a dorsal like k, g, k’, ', x, or χ, which are extremely
infrequent derivatives of a coronal consonant, appears in the series. Moreover, t itself
occurs unchanged in 98 words out of 184, or 53.3%.
Brave new words
If one then compares the set of sounds reecting t- to that of sounds reecting the
nal velar -k, one observes that they are mutually exclusive. Nearly all sounds reect-
ing -k are dorsal consonants like k itself. Dorsals constitute another broad class of
sounds pronounced with the back of the tongue against or close to the hard or so pal-
ates (palatals and velars, respectively) or the uvula (uvulars). All are known to reect
earlier k’s in numerous languages. e only exception is the postalveolar coronal č
(with 6 occurrences reecting -k, or 3.3%), which is a frequent outcome of a former k
in the vicinity of an i or an e (e.g. Latin civitatem [kiwitate] ‘city’ > Italian città [čitta],
or centum [kentu] ‘hundred’ > Italian cento [čento]). And k itself occurs unchanged in
97 words, or 52.7% of the total.
Obviously, the number of individual sounds reecting each original consonant
ought to be related to the number of phonemes that do not reect this sound. And
this relationship is easy to establish. No need to investigate the phonetic inventories
of all the 1,317 languages counted by Boë et al. In their 27 etymologies, Bengtson and
Ruhlen have used a clear principle: potential reexes of a consonant essentially fall into
six categories dened by their point of articulation: with the lips (labials), the tip of the
tongue (coronals), and its back (dorsals). ese articulatory features, which are among
the most resistant in phonetic evolution, combine with the opposition oral-nasal, also
very resistant to change. us, every consonant in a word has on average 1 chance
out of 6 of falling into any of the six categories: oral labial, oral coronal, oral dorsal,
nasal labial, nasal coronal, or nasal dorsal. For a two-consonant root like tik, there is
(1/6)×(1/6) = 1/36 chances that its two consonants will each fall into a particular cat-
egory. And, in any given language, any two-consonant word root thus has 1 chance out
Table 2. Number of occurrences of each reex sound in the 184 presumed cognates sup-
porting Bengtson and Ruhlen’s (1994) Proto-Sapiens series tik ‘nger, one’
Tt 98
ɬ 3
d 23
t′ 2
ts 13
tś 1
s 13
ć 1
č 6
ṭ 1
ɗ 4
dl 1
z 4
š 1
tl 4
,- 1
th 3
tl′ 1 ts′ 3
ř 1
I
i 56 (ī 5) e 33 (ē 5) a 22 (ā 1) o 18 (ō 4) u 10 (ū 1) (monophthongs,
# 2 (#c 1) / 2 0 2 ə 2 ö 1 y 1 with long vowels between parentheses)
ai 3 ia 2 ay 2 ei 1 yi 1 a/ 1 i1e 1 ea 1 oe 1 (diphthongs)
Ø (zero) 5 (loss)
K
k 97 g 14 ʔ 10 ŋ 7 č 6 h 5 q 4 kk 4 k′ 4 x 3kw 2
ʔk 2 nk 2 c 2 ' 1 kp 1 gb 1 kh 1 qw 1 h′k 1 xk 1 jj 1
Ø (zero) 14 (loss)
For each of the three sounds t, i, and k, the assumed reex sounds have been counted. e relatively
numerous sounds reecting each of the original consonants t and k constitute mutually exclusive
sets (with the sole exception of č, as it is a likely derivative of both t and k, particularly in the vicinity
of ani). Vowels are much less stable in all languages, and the assumed reexes of i cover the whole
spectrum of vowel qualities; nevertheless, high front vowels close to i (i, /, e, #) and diphthongs with
an i or an e make up an overwhelming majority of the total (117 out of 184, or 63.6%).
Pierre J. Bancel & Alain Matthey de l’Etang
of 36 – or 0.028, a tiny probability indeed – that each of its consonants will fall within
a particular category.
is parameter can be calculated correctly aer all. And it shows that Bengtson
and Ruhlen’s alleged phonetic laxity is a strong constraint imposed on the discovery
of potential reexes.7
.. Inaccuracy with regard to semantic correspondences
Boë et al. nally nd that Bengtson and Ruhlen are lax with regard to meanings as
well. And this assessment appears to be just as accurate as that regarding sounds: the
apparent variety is great, but the actual diversity is small. Let us again examine how
the various meanings of the words reecting Proto-Sapiens tik ‘nger, one’ quoted by
Bengtson and Ruhlen are represented in their data (Table 3).
Table 3. Number of occurrences of each of the 30 dierent meanings in the presumed
cognates supporting Bengtson and Ruhlen’s (1994) Proto-Sapiens series tik ‘nger, one’
‘on e’ 67 ‘nger’ 37 ‘hand’ 23 arm 10 ‘ten’ 9
‘to show, point’ 5 ‘toe’ 5 ‘only’ 5 ‘ve’ 4 ‘alone’ 4
‘index nger’ 2 ‘middle nger’ 2 ‘only one’ 2 ‘ngernail’ 2 ‘thing’ 2
‘rst’ 1 ‘to say’ 1 ‘one by one’ 1 ‘thumb’ 1 ‘once’ 1
‘foot’ 1 ‘with the ngers’ 1 ‘in hand’ 1 ‘to carry in hand’ 1 ‘by ones’ 1
‘paw’ 1 ‘single’ 1 ‘forenger’ 1 ‘palm of hand’ 1 ‘guy’ 1
Total number of occurrences = 194 (> 184 because of a dozen words with two meanings).
Here again, 30 dierent meanings are represented in Bengtson and Ruhlen’s series.
But a glance at the number of occurrences of each meaning in their sample immedi-
ately shows that the two main meanings, namely ‘nger’ and ‘one’, which are closely
linked together by the universal habit of counting on one’s ngers, account for 104 of
the 194 total meanings, or 53.6%.
e other, less-represented meanings should not be counted as weakening the
numerous convergent words meaning ‘nger’ or ‘one’ – or Bengtson and Ruhlen could
simply have not included them in their series in the rst place, just as they did not
include words meaning ‘elephant’ or ‘carmagnole’, even if they might have t pho-
netically. ough coherent with the two basic meanings from a historical viewpoint,
. We did not take into account the fact that consonant devoicing is respected in 79.9% of
sounds reflecting t- and in 73.9% of those reflecting -k, nor of the fact that 63.6% of vowels
are close phonetic images of -i-; these non-exclusive features are more difficult to integrate,
but may only have a further strong restrictive effect on the probability that the series might
have emerged randomly.
Brave new words
words meaning ‘hand’, ‘ve’, ‘once’, etc. represent a bonus, oen powerful when they
are known to descend from an original word with one of the two critical meanings in
their low-level family.
Moreover, the validity of an etymological meaning does not depend only or even
primarily on the number of attestations of each modern meaning reecting it, but
much more on the reconstruction of a semantic evolutionary process. e original
meaning of a word may have survived in few or even none of its descendants, while
derived meanings may have proliferated. Obviously, Bengtson and Ruhlen’s tik series
would have been much weaker if each of the 30 dierent meanings in their 184-word
sample had been represented by 6 or 7 words, distributed without any evolutionary
logic over the 21 families where reexes of tik are found, contrary to what may readily
be observed in the sample of Indo-European reexes of tik in AppendixG.
e probability of nding a root with an initial t- (or any other oral coronal) fol-
lowed by a -k (or any other oral dorsal) with either of the two meanings ‘nger’ or ‘one’
is double of that of nding a phonetically tting word with only one particular mean-
ing. As a result, 1 language out of 18 (instead of 36) should display consonants from
two particular sets in a word with one of the two meanings ‘nger’ or ‘one’ by the eect
of chance. is probability of 1/18, or 0.056, is still low, and it should apply, following
Boë’s method, to all 104 languages where words with one of these two meanings have
been found. (But we have seen in Section 2.2.1 above that these 104 languages are far
from being the only ones to take into account, and, moreover, that their number is not
really relevant.)
e 90 words with other meanings should be given a higher probability, though
certainly not of 1, depending on the number of evolutionary steps separating them
from the original meaning and on the number of words likely to be reached at each
step. But calculating their respective probabilities, for each word in each language,
would require very long investigations, which are not necessary with Bengtson and
Ruhlen’s etymologies. In the 21 families where they found it – out of their 32 low- or
medium-level language families covering all existing languages – tik must have had
‘nger’ or ‘one’ as its etymological meaning in at least Niger-Congo, Nilo-Saharan,
Afroasiatic, Uralic, Korean, Eskimo-Aleut, Yeniseian, Sino-Tibetan, Na-Dené, Miao-
Yao, Daic, and Amerind, to which one can likely add Indo-European and Turkic. To
retain only the most secure ones, there are 12 ancestral languages displaying a root
meaning ‘nger’ or ‘one’ with an initial coronal and a nal dorsal consonant, a pho-
netic conguration which should occur by chance in 1 language (or ancestor language)
out of 18 – not in 12 out of 32. e actual presence of tik-type roots with secure mean-
ings ‘nger’ or ‘one’ in 37.5% of the world’s language families is thus at least 6.8 times
above the 5.6% chance level. And this gap between chance and facts could only be
enhanced, though more modestly, by the 9 other families with less strong semantic
correspondences.
Pierre J. Bancel & Alain Matthey de l’Etang
In short, counting the number of dierent meanings reecting an original mean-
ing in order to assess the plausibility of an etymological series is, strictly speaking,
meaningless. For each word reecting the proposed root in a given language, the plau-
sibility of its semantic derivation (if any) must be assessed in the light of related words
in its family as well as in closely related families. In Appendix G1, we can see that the
PIE root *deik’e- ‘to point’ has descendants endowed with verb meanings as dierent as
‘to point out, to show, to exhibit, to confess, to say, to teach, to accuse, to manifest, to
give a sign’ and others, plus nouns as disparate as ‘direction, region, part, earth, world,
camping ground, country, village, cultivated eld, side, span, hand span, amazement,
nger, toe, accusation, sign, example, token, dedicace, discourse’, and ‘judge’, total-
ing 31 dierent meanings (and more could be added). Is PIE *deik’e- disqualied by
this variety? Certainly not, because the variety is only supercial, and in each Indo-
European subgroup meanings are organized into apparent logical evolutionary chains.
is evolutionary logic cannot be adequately accounted for by a statistical model.
.. Summary
– e negative conclusions of the probabilistic calculations we have examined (Boë
2004; Boë et al. 2003, 2006; Ringe 2002) cannot be regarded as valid.
– Although it seems relatively easy to take into account the degree of phonetic
validity of assumed reex words, it is very dicult to reduce to gures the dif-
ferences in taxonomic level between languages (the greater etymological weight
of, e.g. Proto-Indo-European against any of its descendants, or of Basque against
Gascon), or in logically derived meanings in a linguistic lineage versus meanings
picked up at random without regard to semantic evolutionary logic (e.g. the logi-
cal validity of deriving ‘toe’ from ‘nger’, against the invalidity of directly deriving
‘toe’ from ‘to point’). More work will be necessary to perhaps achieve a satisfac-
tory assessment of etymological series by mathematical means.
– A point that is relatively dicult to conceive and understand is how multilateral
etymological series dier from phonetically regular etymological series in lower-
level language families. e latter (like those shown in Appendices G1 and G2, PIE
*deik’e- ‘to point, to show’ and *dekm) - ‘ten’, respectively) aim to trace with certainty
the descent of a root in all the descendant languages. e most powerful tool to
ascertain that words from dierent languages belong to such etymologies is regular
phonetic correspondences, which may practically eliminate any doubt that a partic-
ular word displaying them belongs to a given series, without any probabilistic assess-
ment being needed – not because there is any magic in regular correspondences,
but because they link together dozens of word series by their constituent sounds in
metaseries whose appearance by chance would obviously have been highly improb-
able, just as no calculation is needed to realize that, say, getting 200 aces of hearts
when taking a card at random from each of 200 decks is a near impossibility.
Brave new words
Multilateral series, in turn, rely on phonetic correspondences that are oen not
demonstrably regular in the state of our knowledge; in other words, they are not found
again and again across dierent word series. But the phonetic nature of these corre-
spondences otherwise complies, within each series, with evolutionary rules that have
been discovered, over the last two centuries, in low-level families thanks to regular
correspondences. As we have seen in Section 2.2.2, these rules impose a strong con-
straint on the discovery of potential cognates. is constraint is, however, weaker than
that posed by regular correspondences themselves and does not warrant that each
particular word included in a series really belongs to it; nevertheless, if many words
in a series repeatedly satisfy this constraint, the likelihood that the entire series has
appeared at random quickly drops. Consequently, a multilateral series warrants the
authenticity of a root in a proto-language, while none of its assumed descendants may
be considered to descend from it with perfect certainty – even if, taken collectively,
most of them must descend from it.
is apparent paradox was expressed by Bengtson and Ruhlen:
We do not harbor illusions … that every etymological connection we propose
will be found, ultimately, to be correct, but we do believe that the removal of such
errors as may exist in these global etymologies will not seriously aect the basic
hypothesis, which does not depend on any specic link for its validity.
(Bengtson & Ruhlen 1994, p. 292)
What replaces regular phonetic correspondences in multilateral series is the number
of families involved in them, and the recurrence of series within a particular group of
families, such as that of *m- ‘rst person’ and *t- ‘second person’, which are (together
with many others) particular to the group of families Greenberg (2000–2001) calls
Eurasiatic.
– Many etymologies presented by Greenberg and Bengtson and Ruhlen, including
the ones discussed above, are so massively supported that no probabilistic calcu-
lation is needed. Just like papa/mama words or, for that matter, reconstructions
supported by regular sound correspondences, they are far beyond the point where
sophisticated tools might be necessary.
– However, accurate probabilities might be useful to uncover other, less well-
preserved roots, to assess disputed taxa, and more generally to enlarge our com-
parative knowledge of remote language families. One can only encourage both
statisticians and comparatists to continue to address this dicult problem in a
constructive spirit.
– If papa/mama words have managed to last for several dozen millennia, why could
not some other words have resisted as well? And perhaps not so few of them – aer
all, Bengtson and Ruhlen’s 27 Proto-Sapiens etymologies result from the eorts of
two scholars, while hundreds of Indo-Europeanists have worked over the two last
centuries on a few dozen closely related languages to unearth some 2,500 PIE roots.
Pierre J. Bancel & Alain Matthey de l’Etang
. Why kinship appellatives do not change: Children babbling, parents
choosing
Let us now examine two lines of evidence from dierent elds of the study of language,
which converge with our own to support the hypothesis that papa/mama words must
have played a crucial role in the early appearance of articulate speech.
Papa/mama words have been preserved over the whole history of language fami-
lies with a written tradition, as documented in Section 1 for a number of such families,
and comparison within language families with no written record shows that such is
the case for them as well (Matthey de l’Etang & Bancel in preparation). Why is it, then,
that they are not – or, at least, very infrequently – subject to phonetic change and word
replacement, as all other words are? e reason is simple and compelling, and every
parent who has raised a child who developed normal speech knows it, but this com-
mon experience has percolated into the domain of scientic knowledge only recently
and without attracting much attention. Papa/mama words are crucial for babies to
learn and for parents to teach babies to speak. e actual mode of their transmission
has been explained by the language acquisition specialist John Locke (1990), and it is
a nice piece of collaboration between parents and children.
Around the age of 6 to 9 months, on average, all babies enter the babbling stage
of language acquisition. Canonical babbling consists of repetitive bababa, papapa,
mamama, dadada, tatata, and nanana syllables, made up of plain labial or coronal
nasal or oral stops, plus an open vowel (Oller 1980). It has long been recognized that
these syllables are the rst to be mastered by children because they are the easiest,
due to a range of constraints (Westermarck 1891; Jakobson 1960; MacNeilage & Davis
1990; MacNeilage 2008).
Among these sequences, parents “recognize” those corresponding to a word in
their language and reinforce them – notably by repeating them in their standard form
while pointing a nger at the parent concerned – while they leave unreinforced other
sequences that do not match with a word in their language, and which the child will
thus progressively abandon.
is was Locke’s (1990) great discovery, which denitively falsies the theory of
the spontaneous emergence of these words. Or, rather, it falsies the theory that babies
invent them alone. Children spontaneously provide a range of syllabic frameworks,
and parents rectify some of them into the canonical forms of the corresponding words
in their language: English parents reinforce dadada and mamama into dad and mom
(or mum), respectively; French parents reinforce papapa and mamama into papa
and maman, respectively; Turkish parents reinforce bababa and nanana into baba
and anne (a word related to Proto-Turkic ana ‘mother’, also inherited in Turkish; see
Appendix E3); and so on. It would never occur to a monolingual English mother to
induce her daughter to call her anne (even if her own given name is Ann or Annie),
Brave new words
nor to monolingual French parents to recognize in their baby’s babbling of dadada a
word meaning ‘dad’ and to reinforce it.
is crucial way of transmitting papa/mama words explains why English children
consistently learn dad and mom, French children papa and maman, Turkish children
baba and anne, Hindi-speaking children bap and mã, and so on. Each of these words
belongs to the lexicon of a particular language. Children provide the initial spontane-
ous syllabic framework; the exact phonetic form and meaning of each word are taught
by parents. is fact, which clearly implies lexical inheritance rather than innovation,
was elusively recognized by Jakobson (1960) in the paper in which he paradoxically
argued for spontaneous innovations instead of common descent:
[C]hildren, being prompted and instigated by the extant nursery words, gradually
turn the nasal interjection into a parental term and adapt its expressive make-up
to their regular phonemic pattern.
is “prompting and instigation by extant nursery words” discreetly acknowledges
the fact that parents reinforce their child’s babbling and shape it into already existing
words. And behind this teaching stands an uninterrupted transmission from genera-
tion to generation.
is specic mode of transmission also explains why these words change so
rarely. When a language is in the process of undergoing a phonetic change that
should change their form – for instance when stops between vowels change to frica-
tives (a very common type of change), so that baba, papa, dada, and tata should
become bava, pafa, daza, and tasa, respectively –, the bio- and neuromechanical
constraints bearing on babies who are learning to speak at that particular time are
most of the time stronger. Babies do not master fricatives and continue to say baba,
papa, etc. preventing the change from applying to the word in question; parents
recognize the form baba or papa they have heard since their own childhood and
reinforce it rather than the modied form, which in any case exceeds the baby’s
articulatory capacities.
As a result, these words are transmitted from one generation to another without
change, and are unlikely to be lost, since the same spontaneous syllabic frameworks
reappear every time another child reaches the age of 6 to 9 months and begins bab-
bling – a phenomenon which must have occurred regularly in all human groups that
have survived long enough for us to know something of their language, and thus have
covered nearly all periods of phonetic change in all languages.
Finally, papa/mama words are crucial in another aspect of language transmission.
In children’s rst utterances, they function no dierently from animal communica-
tion. ey have been dubbed holophrastic words (“a whole phrase in one word”; see
De Laguna 1927), because they seem to convey information that should be rendered
in adult language by a complex sentence. Brigaudiot and Danon-Boileau (2002), in a
Pierre J. Bancel & Alain Matthey de l’Etang
section entitled “Les premiers maman, holophrases ou énoncés à un terme” [e rst
maman, holophrases or single-term utterances], quote a century-old analysis:
Childish mama, translated into advanced speech, does not mean ‘mother’ but
rather a sentence such as ‘Mama, come here’, ‘Mama, give me…’, or ‘Mama, put
me in the chair’, or ‘Mama, help me’. (Stern & Stern 1907)
ese holophrases are similar to the calls of young animals “holophrastically” call-
ing their mothers, except that the human baby’s call, contrary to those of all other
animals, is phonetically articulate: it consists of vowels and consonants arranged into
syllables. But papa/mama words do not remain mere calls for long. Within a few weeks
or months, reinforcement by elders, together with the recurrence of the association,
in the parents’ speech, of one particular reinforced sound sequence with the presence
of the mother, and of another one with the father, induces the child to establish a link
between each of these sequences and a particular being in the outside world. And this
association is crucial, since it opens the door of symbolic meaning for the child.
In this way, too, parental appellatives play a unique role in the transmission of
language. And it must have been so for untold ages.
. Back to Proto-Human: e Frame, then Content hypothesis
Papa/mama words have survived – or, rather, their continuous transmission and pres-
ervation was necessary to our ancestors – over the last 2,000 to 10,000 generations.
During this period, they have been crucial for babies learning to speak – and their par-
ents teaching them – in the nice collaborative eort described by Locke. Why should
they not have been preserved over the 20,000 to 100,000 generations before that? We
suggested long ago that kinship appellatives might have been among the very rst pho-
netically articulate sounds (Bancel & Matthey de l’Etang 2002), no doubt a long time
before Proto-Sapiens was spoken.
At that time, we were not aware of Peter MacNeilage’s (2008; see also MacNeilage,
this volume) “Frame, then Content” phonetic theory of the origin of speech, rst pre-
sented with respect to modern babies by MacNeilage and Davis (1990). is theory
holds that papa/mama words must be the rst sound sequences mastered by a human
mouth, for compelling phonetic reasons discovered through the observation of lan-
guage acquisition. To understand these reasons, one has to recall that all humans
speaking a language, whatever their individual dierences, are true virtuosos – just
like all falcons are nonesuch sky-divers, or all whales are outstanding apnea sea-divers,
as a result of major selective pressures.
As explained by Lieberman (1992) – whose pioneering work (e.g. Lieberman
1975, 1985, 2000) opened the door to the study of language evolution, which had
Brave new words
remained barred for a century – speaking is the most dicult motor activity, because
of the extreme speed and precision of the successive motions involved in the articu-
lation of a speech sound string. According to MacNeilage (2008), about 40 dierent
muscles are involved in the production of the various speech sounds, each perform-
ing a very dierent function such as controlling the pressure of the airow breathed
out of the lungs, or the tension of vocal cords, opening and closing the nasal airway,
or giving the vocal tract a particular shape. Based on an average 15 of these muscles
being involved in each particular sound, and a speech rate of 15 sounds per second,
MacNeilage arrives at the fantastic number of 225 muscular actions per second in
speech, or one every 5milliseconds. Most of them must be eected with millimetric
precision, and all must be tightly coordinated; otherwise the sounds produced are not
those intended. Such defects in coordination do indeed happen and are a major source
of phonetic evolution, showing that when speaking we are always at the extreme limit
of our capacities, without even being aware of it.
On the auditory side, the high speed of some 15 to 25 units per second at which
speech sounds are normally delivered is equally amazing. Hearers decode them easily,
although it is oen beyond the speed limit of 15 units per second beyond which other
sounds merge into an undierentiated buzz in the hearer’s perception. And the brain
areas and connections able to process this high-speed auditory ow can do so e-
ciently only aer appropriate training – that is, learning the language. Just think how
dicult it is, when you start learning a new language, merely to perceive the sounds
you are not used to.
e extreme diculty of both speaking and hearing an articulate language may
be the reason why babies spontaneously start babbling in the second half of their rst
year. is universal behavior must rely at least partly on an innate trend, resulting from
a heavy selective pressure exerted on humans to begin speaking at an early age, so they
can gain the required uency, again conrming that articulate speech has long been a
major feature of the human ecological niche.
It also explains why babbling consists of plain stops and vowels in the simplest
syllable sequences. Babbling typically reduplicates the most basic articulate syllables,
namely consonant-vowel (CV), in CVCVCV… sequences. ese syllable sequences
using only two sounds are the easiest way to produce an articulate speech ow, as they
require the synchronization of very few muscles into a repetitive, dual motor scheme,
however long the syllable sequence may be. Moreover, MacNeilage and Davis (1990)
found in early babbling an inertial pattern whereby the tongue stays in the same posi-
tion for the vowel as it was in for the previous consonant, or the surrounding conso-
nants in reduplicative babbling. Consequently, to produce bababa, an infant initially
needs only a couple of mandible elevation muscles and a couple of mandible depres-
sion muscles. For dædædæ, she only needs to add the inferior genioglossus to move the
tongue forward and up, and for gogogo only two or three muscles are added to the ones
Pierre J. Bancel & Alain Matthey de l’Etang
used for mandibular oscillation (Peter MacNeilage, personal communication).8 From
the neuromotor viewpoint, this is a huge simplication with regard to the require-
ments of adult speech.
e complete closure of stops also allows much more variation in the articulatory
motions than for any other speech sound. No matter what the speed, strength, and
precision of the closing motion to produce a stop may be – whichever way the airow
is closed and reopened, it will produce an acceptable approximation of the intended
sound. In contrast, other consonants such as fricatives or glides demand millimetric
precision in their execution, and any deviation from the intended target is likely to
drastically modify the acoustic output.
Furthermore, as already noted by MacNeilage and Davis (1990), a babbling
sequence essentially relies on motions that lower and raise the jaw – a motion over
which voluntary control has been selected in the human lineage since our distant Gna-
thostomata ancestors, which appeared some 450million years ago, acquired a mouth
with a jaw.
e articulatory, motor, and syllabic robustness of consonants p, b, m, t, d, n is
the reason why these speech sounds are the rst ones children regularly master in the
articulated syllable sequences papapa, bababa, mamama, etc. Of course, if one ran-
domly “tries” one’s articulatory organs in order to make a sound, any human phoneme
(and many other sounds) may result. However, when it comes to reproducing a sound,
and – which is still more dicult – a sequence of two sounds at will, of course the easi-
est sounds and sequences must be the rst to be mastered.
is is exactly what children do, and what humans learning to speak – either early
in life or at an older age – must always have done. Both MacNeilage (2008) and we
(Bancel & Matthey de l’Etang 2002, 2005) independently arrived at the conclusion that
this rule must have been in force since the very beginning of articulate speech – not
only in Homo sapiens, but in our more ancient human ancestors as well.
. Our comparative data converge with MacNeilage and Davis’ (1990) finding concerning
the detail of vowels in early babbling. According to them, children’s first velar consonants
occur with a velarized vowel in sequences like gogo or koko. While compiling our kaka etymo-
logical series, we were soon struck by finding a high number of koko ~ kuku (or gogo ~ gugu)
forms, sometimes even predominant over kaka forms, as in Nilotic or Southern Amerind, but
also occurring sporadically in many other language groups. In contrast, popo ~ bobo and toto
~ dodo variants of papa ~ baba and tata ~ dada are extremely rare. We could not find any
consistent semantic correlate of this variation in the vowel. MacNeilage and Davis’ finding
regarding modern children may be regarded as confirmed by this globally frequent variant.
Reciprocally, while it does not help to resolve the question of the antiquity of these koko ~
kuku variants, it provides an explanation for their existence.
Brave new words
. By way of conclusion: e early steps towards articulate language
As we have seen, there are three lines of independent ndings. e rst nding is
that of MacNeilage and ourselves, based on the phonetics of language acquisition, that
papa/mama sound sequences are the obligatory rst steps towards mastering articu-
late speech, and must have been so throughout human history. e second one is that
of Locke, showing how children and parents cooperate in the transmission of papa/
mama words; even if Locke himself does not consider the issue from an evolutionary
perspective, there is no doubt that this mode of transmission is not recent in humans.
And the third is Ruhlen’s and our own nding, supported by data from thousands of
languages worldwide, that most papa/mama words can only have been inherited from
a common Proto-Sapiens language. All three lines of evidence converge on a scenario
in which kinship appellatives must have early played a prominent role in the evolution
of speech in humans and might even have been at its very origin.
Beyond this striking convergence, this scenario has other aspects adding to its
evolutionary value. In particular, the initial acquisition by babies of phonetic articula-
tion in their babbling stage through meaningless syllable sequences, some of which are
then given a meaning by parents, seems to be a step towards the solution of a mystery
that has barely been noted, much less explained, since research about language origins
has burgeoned.
Words have to have been invented, however long this invention may have taken.9
But how? Both phonetic articulation and referential meaning are unprecedented in
animal history, and both are too complicated to have been developed simultaneously.
e rst step towards the elucidation of their origin must therefore be to discover
which appeared rst. Babbling babies show us that phonetic articulation appears rst
in all contemporary individuals. And it must have been so originally as well, since
speech is such a dicult activity that, if humans had found another way to convey
referential meanings in the beginning, they certainly would not have gone to the trou-
ble of trying to move their tongues and lips at an incredible speed from one incred-
ibly weird position to another but would have stuck to the previously used means of
expression and developed it further. Articulate speech must have been discovered by
chance, as was the case with all biological evolutions before and aer it, and in its sim-
plest form – that of mamama, papapa, bababa sound sequences. It must also have been
initially used to fulll previously existing communicative, non-referential functions.
Only later, probably much later, did its wonderful but highly demanding properties
. e reluctance to deal with the emergence of words is most conspicuous in Kenneally’s
(2007) book e First Word. In spite of its title, this summary of the current state of research
about language evolution does not even allude to the question.
Pierre J. Bancel & Alain Matthey de l’Etang
allow for a very slow dierentiation of sequences based on very few consonants. It
opened the door to a functional dierentiation, which, ultimately, led to the emer-
gence of semantic reference.
e consonants in kinship appellatives already delineate a simple phonetic feature
system, based on articulatory motions and the corresponding bundles of neuromotor
commands, each of which must be called into play with dierent command bundles
to produce dierent consonant sounds. Appellatives also constitute a simple seman-
tic system, based on a few obvious semantic features, the rst being the opposition
between males and females. ey thus oer a plausible path to the development of
structured phonetic and semantic systems, whose interrelated features have made us
the “symbolic species” (Deacon 1997, p. 87, Figure 3.3).
Finally, let us allude to the fact that kinship is another uniquely human trait,
whose insertion in the humanization evolutionary process has hardly been discussed
before, in spite of the many promising avenues it oers. Articulate language, this
essentially social human ability, might not have developed without a reinforcement of
social bonds, and kinship has long been the primary mode of human social organiza-
tion. e antiquity of kinship is warranted by both the universality of kinship systems
in all known human groups and the existence of precursors of kinship relationships in
apes. Given the complexity of both language and kinship, it is only natural that they
have coevolved, further enhancing the plausibility that the rst symbolic meanings
ever acquired by humans concerned kinship relations.
. How else may Proto-Sapiens aid the study of language origins?
Finally, let us illustrate briey how remote etymologies could shed light on other
aspects of the evolution of language ability. Apart from papa/mama words, the most
resistant words worldwide are rst- and second-person pronouns (Dolgopolsky 1964;
Pagel 2000). In all families, they display an incredible resistance, as compared to the
average replacement rates of 13% to 18% per millennium that have been calculated
for the 100 or 200 most basic words (body parts, natural elements, kinship relations,
pronouns, basic verbs, etc.).
In an unpublished study bearing on 494 Indo-European languages and dialects,
we have found that the PIE rst-person *m- and second-person *t- have been lost, aer
6,000 to 8,000 years, in an amazingly small number of descendant languages. First-
person *m- was lost by only two languages (0.4%), which amounts to a loss rate of
0.05% per millennium, granting *m- a half-life of 1.38 million years.10 In turn, *t- has
. e calculation of the half-life of words was devised by Pagel (2000). It is not as reliable
as its prototype in physics, where one observes the decay of a given quantity of an element
Brave new words
disappeared from seven languages (1.4%), which endows it with a loss rate per millen-
nium of 0.18%, and a half-life of 385,000 years.
Personal pronouns from most language families display similarly minuscule loss
rates. However, unlike papa/mama words, and contrary to what might be expected
given this extraordinary longevity, there is no global convergence of phonetic forms
and meanings in rst- and second-person pronouns. We have studied (Bancel & Mat-
they de l’Etang 2008, 2010) the phonetic distribution of pronoun roots in shallow-time
ancestral pronominal forms worldwide compiled by Ruhlen (1994b, pp. 252–260) –
who, interestingly, did not discover any Proto-Sapiens rst- or second-person pronoun
root in spite of the pleasure he no doubt would have had in nding one. We have found
that a majority of these pronoun roots are based on a handful of consonants, which,
however, are distributed among the rst and second persons in apparent disorder at
the global level. A root m- may represent the rst person singular in some phyla (like
Eurasiatic or Niger-Congo), or the second person singular in others (like Amerind),
and the same holds true of the other globally widespread pronominal root consonants
t-, n-, k-, and s-, in spite of their monolithic coherence at the family-internal level.
Another salient aspect of the phonetic distribution of pronominal root conso-
nants is the near absence of plain oral labial stops (p-, b-), with very few exceptions,
and those few are oen demonstrably secondary, such as bi- ~ be- ‘I (nominative)’ in
Altaic languages.11 While this global absence remains unexplained, its very existence
must be considered as indicative of a relationship between all pronominal forms. Given
that plain oral labial stops are among the most frequent consonants in the world’s lan-
guages (Maddieson 1984, 1997) and are rather resistant to phonetic change, if rst-
and second-person pronouns had independent origins in many language families, a
good number of them ought to be based on a root p- or b-.
We then thought that rst- and second-person pronouns (and rst- and second-
person markers more generally) may have emerged only with the uent use of syntac-
tic articulation, and the necessity to quickly dierentiate the speaker and the hearer in
a complex sentence. In the stages that preceded the evolution of syntactic articulation
in a broad sense – stringing words together – words were mostly used in isolation,
and a great proportion of the speaker’s intentions had to be inferred from the context.
Words were, however, highly useful, as compared with no words at all, thanks to their
over time. With words, one may only observe their loss as the ancestral language splits into
multiple descendants. It does, however, give a good indication of their relative stability.
. e Altaic language family consists of Turkic (Turkish, Uzbek, Kazakh, etc.), Mongolic
(Classical Mongolian, Khalkha, Buriat, etc.), and Tungusic (Manchu, Evenk, Nanai, etc.);
Korean and Japonic, thought by Greenberg to be related to the former groups at a greater
remove within Eurasiatic, are oen included within Altaic by Nostraticists.
Pierre J. Bancel & Alain Matthey de l’Etang
property of referring to objects or actions known to the speaker and the hearer. ey
gave the hearer an anchor to infer the rest, in a world where human activities and
interests were much more restricted and predictable than in any society known to us
today. But rst- and second-person pronouns have the strange and unique property of
shiing reference with the speech turn. One does not see how such words, deprived
of the essential quality of words at that time of referring to a stable object, could ever
have appeared, nor what use they might have had – a single-word sentence “me” or
“thou” would have given little information to the hearer. When syntactic articulation
rst began, verbs may only have been “action or state words,” with no mark of tense,
voice, person, or number, just as they had been before, when used in isolation. As
stringing words together became a widespread habit, then a norm, it became neces-
sary to disambiguate the subject and object of verbs – very oen the speaker or hearer
themselves, the most interesting themes for two interacting individuals – with nouns
used to address the hearer and self-refer to the speaker. anks to this repetitive use,
the most frequent of these nouns must, by a process which remains unclear (although
probably not forever), have evolved in shortened forms into rst- and second-person
pronouns.
Our conclusion was that, at the time of Proto-Sapiens, personal pronouns were
already being formed, since they are found in all language families (in spite of their
not being absolutely necessary, albeit very useful) but were not yet xed as a lexical
category. e original nouns that had given rise to them still varied freely between
referring to the speaker and referring to the hearer, according to their original nominal
meaning, and only later were xed onto either rst or second person in each family.
Since the very existence of rst- and second-person pronouns is hardly conceivable
without a syntactically articulated language, Proto-Sapiens at the time of its split must
have been in the process of acquiring syntactic articulation. is process certainly took
time, and perhaps lasted until late into Upper Paleolithic, judging by the fact that there
are more reconstructed rst- and second-person pronoun roots in ancient taxa, such
as Eurasiatic (Greenberg 2000) or Nostratic (Bomhard 2008; Dolgopolsky 2008), than
in recent ones. e unexpected absence of a clear-cut distinction between rst- and
second-person pronominal roots at the global level would thus testify that syntactic
articulation had begun to evolve before the dispersion of modern humans, and prob-
ably was part of its success, but had not yet led to the development of full-edged rst-
and second-person pronouns.
It has been repeated recently that the origin of language is the most dicult sci-
entic problem of our time. At the very least, it is certainly the most dicult problem
resisting evolutionary theory. How could one hope to solve it without the powerful
tool of comparative linguistics, which opens a window on past times as far back as
the initial dispersion of our Homo sapiens ancestors? How could one hope to solve it
without giving spoken words their legitimate due?
Brave new words
References
Abaev, V.I. (1970). Osetinsko-Russkij Slovar’ – Iron-Uyryssag Dzyrduat. Ordžonikidze: IR.
〈http://ironau.ru/lingvo-abbyy.html〉 (31May 2010).
Academia Română. (1998). Dicţionarul explicativ al limbii române. Bucharest: Univers Enciclo-
pedic. 〈http://dexonline.ro/〉 (15 April 2010).
Bancel, P.J., & Matthey de l’Etang, A. (2002). Tracing the ancestral kinship system: e global
etymon kaka. Part I: A linguistic study. Mother Tongue, 7, 209–243.
Bancel, P.J., & Matthey de l’Etang, A. (2005). Kin tongue. A study of kin nursery terms in relation
to language acquisition, with a historical and evolutionary perspective. Mother Tongue, 9,
171–190.
Bancel, P.J., & Matthey de l’Etang, A. (2008). e millennial persistence of Indo-European and
Eurasiatic pronouns and the origin of nominals. In J.D. Bengtson (Ed.), In hot pursuit of
language in prehistory. Papers in the four elds of anthropology in honor of Harold C. Fleming
(pp. 439–464). Amsterdam: John Benjamins.
Bancel, P.J., & Matthey de l’Etang, A. (2010). Where do personal pronouns come from? Вопросы
языкового родства /Journal of Language Relationship, 3, 127–152.
Bancel, P.J., Matthey de l’Etang, A., & Bengtson, J.D. (2010). Back to Proto-Sapiens (part 2). e
global kinship terms papa, mama and kaka. In D. Jones & B. Milicic (Eds.), Kinship, lan-
guage, and prehistory. Per Hage and the renaissance in kinship studies (pp. 38–45). Salt Lake
City, UT: University of Utah Press.
Bengtson, J.D. (Ed.). (2008). In hot pursuit of language in prehistory. Papers in the four elds of
anthropology in honor of Harold C. Fleming. Amsterdam: John Benjamins.
Bengtson, J.D., & Ruhlen, M. (1994). Global etymologies. In M. Ruhlen (Ed.), On the origin of
languages. Studies in linguistic taxonomy (pp. 277–336). Stanford, CA: Stanford University
Press.
Bird, B. (2006). Online Yaghnobi-Tajik-English lexicon. 〈http://yaghnobi.les.wordpress.com/2007/
07/yaghnobi-english-tajik-lexicon.pdf〉 (27 May 2010).
Boë, L. -J. (2004). La langue unique et l’hypothèse de Ruhlen. Repères historiques et mise à l’épreuve
méthodologique. Paper presented to the Société de Linguistique de Paris on March6, 2004.
Boë, L.-J., Bessière, P., Ladjili, N., & Audibert, N. (2006). Simple combinatorial considerations
challenge Ruhlen’s mother tongue theory. In B. L. Davis & K. Zajdo (Eds.), e syllable in
speech production (pp. 63–93). New York, NY: Lawrence Erlbaum Associates.
Boë, L.-J., Bessière, P., & Vallée, N. (2003). When Ruhlen’s “mother tongue” theory meets the
null hypothesis. Proceedings of the 15th International Congress of Phonetic Sciences,
Barcelona.
Bomhard, A.R. (2008). Reconstructing Proto-Nostratic. Comparative phonology, morphology, and
vocabulary. Leiden: Brill.
Bowler, J.M., Johnston, H., Olley, J.M., Prescott, J.R., Roberts, R.G., Shawcross, W., et al. (2003).
New ages for human occupation and climatic change at Lake Mungo, Australia. Nature,
421, 837–840.
Brigaudiot, M., & Danon-Boileau, L. (2002). La naissance du langage dans les deux premières
années. Paris: Presses universitaires de France.
Burrow T., & Emeneau, M.B. (1984). A Dravidian etymological dictionary. (2nd ed.) Oxford, UK:
Clarendon. 〈http://dsal.uchicago.edu/dictionaries/burrow/〉 (5 May 2010).
Buschmann, J.C.E. (1852). Über den Naturlaut. Philologische und historische Abhandlungen der
königlichen Akademie der Wissenschaen in Berlin, 3, 391–423.
Pierre J. Bancel & Alain Matthey de l’Etang
Cann, R.L., Stoneking, M., & Wilson, A.C. (1987). Mitochondrial DNA and human evolution.
Nature, 325, 31–36.
Chantraine, P. (1968). Dictionnaire étymologique de la langue grecque. Histoire des mots.
Paris: Klincksieck. 〈http://www.archive.org/details/Dictionnaire-Etymologique-Grec〉
(7 May 2010).
Charles-Edwards, T.M. (2003). Early Irish and Welsh kinship. Oxford, UK: Oxford University
Press.
Ciorănescu, A. (1958–1966). Dicţionarul etimologic român. Tenerife: Universidad de la Laguna.
Deacon, H.J. (2001). Guide to Klasies River. 〈http://academic.sun.ac.za/archaeology/KRguide
2001.PDF〉 (16 September 2011).
Deacon, T. (1997). e symbolic species. e co-evolution of language and the brain. New York,
NY: Norton.
De Laguna, G. (1927). Speech: Its function and development. New York, NY: Yale University
Press.
d’Errico, F., Henshilwood, C.S., Vanhaeren, M., & van Niekerk, K. (2005). Nassarius kraussianus
shell beads from Blombos Cave: Evidence for symbolic behaviour in the Middle Stone Age.
Journal of Human Evolution, 48, 3–24.
de Waal, F. (1982). Chimpanzee politics. Power and sex among apes. London: Jonathan Cape.
Dolgopolsky, A. (1964). Gipoteza drevnejšego rodstva jazykovyx semej severnoj Evrazii s vero-
jatnostnoj točki zrenija, Voprosy jazykoznanija, 2, 53–63. [English translation: A probabi-
listic hypothesis concerning the oldest relationships among language families of North-
ern Eurasia. In V. V. Shevoroshkin & T. L. Markey (Eds.), Typology, relationships and time
(pp.27–50). Ann Arbor, MI: Karoma, 1986].
Dolgopolsky, A. (2008). Nostratic dictionary. McDonald Institute for Archaeological Research,
University of Cambridge, UK. 〈http://www.dspace.cam.ac.uk/handle/1810/196512〉
(9November2010).
Dybo, A. (2006). Turkic etymology. Online etymological database, Tower of Babel Project.
〈http://starling.rinet.ru/cgi-bin/main.cgi?ags=eygtnnl〉 (26 May 2010).
Emeneau, M.B. (1953). Dravidian kinship terms. Language, 29, 339–353.
Fouts, R.S. (1997). Next of kin. What chimpanzees have taught me about who we are. New York,
NY: William Morrow.
Greenberg, J.H. (2000–2001). Indo-European and its closest relatives. e Eurasiatic language
family. Stanford, CA: Stanford University Press.
Grierson, G. (1920). Ishkashmi, Zebaki and Yazghulami. An account of three Eranian dialects.
London: Royal Asiatic Society. 〈http://www.angelre.com/sd/tajikistanupdate/engpamir-
languages.html〉 (31 May2010).
Hayyim, S. (1934–1936). New Persian-English dictionary. Tehran: Beroukhim. 〈http://dsal.uchi-
cago.edu/dictionaries/hayyim/〉 (4June2010).
Henshilwood, C.S., & d’Errico, F. (2005). Being modern in the Middle Stone Age: Individu-
als and innovation. In C. Gamble & M. Porr (Eds.), e individual hominid in context:
Archaeological investigations of Lower and Middle Palaeolithic landscapes, locales and arte-
facts (pp.244–264). New York, NY: Routledge.
Henshilwood, C.S., d’Errico, F., Marean, C.W., Milo, R.G., & Yates, R. (2001). An early bone tool
industry from the Middle Stone Age at Blombos Cave, South Africa: Implications for the
origins of modern human behaviour, symbolism, and language. Journal of Human Evolu-
tion, 41, 631–678.
Izard, M. (1965). La terminologie de parenté bretonne. L’Homme, 5, 88–100.
Brave new words
Jakobson, R. (1960). Why mama and papa? In B. Kaplan & S. Wapner (Eds.), Perspectives in
psychological theory. Essays in honor of Heinz Werner (pp. 124–134). New York, NY: Inter-
national Universities Press.
Kabir, H., & Akbar, W. (1999). Dictionnaire pashto-français. Paris: L’Asiathèque.
Kenneally, C. (2007). e rst word. e search for the origins of language. New York, NY: Viking.
Keskin, M. (No date). Wörterverzeichnis. Zazaki-deutsch deutsch-zazaki. 〈http://www.zazaki.de/
deutsch/Zazaki-Deutsch_Deutsch-Zazaki_Woerterbuechlein.pdf〉 (31 May 2010).
Knight, A., Underhill, P.A., Mortensen, H.M., Zhivotovsky, L.A., Lin, A.A., Henn, B.M., etal.
(2003). African Y chromosome and mtDNA divergence provides insight into the history of
click languages. Current Biology 13–6, 464–473.
Lieberman, P. (1975). On the origins of language. An introduction to the evolution of speech. New
York, NY: Macmillan.
Lieberman, P. (1985). On the evolution of human syntactic ability. Its pre-adaptive bases: Motor
control and speech. Journal of Human Evolution, 14, 657–668.
Lieberman, P. (1992). On the evolution of human language. In J.A. Hawkins & M. Gell-Mann
(Eds.), e evolution of human languages (pp. 21–47). Redwood City, CA: Addison-Wesley.
Lieberman, P. (2000). Human language and our reptilian brain. e subcortical bases of speech,
syntax, and thought. Cambridge, MA: Harvard University Press.
Locke, J.L. (1990). “Mama” and “papa” in child language. Parent reference or phonetic prefer-
ence? In B. Metuzale-Kangere & H.D. Ringholm (Eds.), Symposium balticum. A festschri
to honour Professor Velta Ruke-Dravina (pp. 267–273). Hamburg: Buske.
Lubbock, J. (1889). e origin of civilisation and the primitive condition of man. Mental and social
condition of savages (5th ed.). London: Longmans & Green.
Lubotsky, A. (No date). Indo-Aryan inherited lexicon. Online etymological database of the
Indo-European Etymological Dictionary project, University of Leiden. 〈http://www.indo-
european.nl/index2.html〉 (20 October 2010) [transferred to Brill’s website].
MacNeilage, P.F. (2008). e origin of speech. Oxford, UK: Oxford University Press.
MacNeilage, P.F., & Davis, B.L. (1990). Acquisition of speech production: Frame, then content.
In M. Jeannerod (Ed.), Attention and performance, 13: Motor representation and control
(pp. 453–476). Hillsdale, NJ: Lawrence Erlbaum Associates.
Maddieson, I. (1984). Patterns of sounds. Cambridge, UK: Cambridge University Press.
Maddieson, I. (1997). Phonetic universals. In W.J. Hardcastle & J. Laver (Eds.), e handbook of
phonetic sciences (pp. 619–639). Oxford: Blackwell.
Mahadevan, I. (2003). Early Tamil epigraphy from the earliest times to the 6th century AD.
Cambridge, MA: Harvard University Press.
Mancino, T. (No date). Shughni dictionary. 〈http://web.linguist.umass.edu/~woolford/
Shughni%20Dictionary.pdf〉 (31 May 2010).
Marean, C.W., Bar-Matthews, M., Bernatchez, J., Fisher, E., Goldberg, P., Herries, A.I.R., etal.
(2007). Early human use of marine resources and pigment in South Africa during the
Middle Pleistocene. Nature, 449, 905–908.
Matthey de l’Etang, A., & Bancel, P.J. (2002). Tracing the ancestral kinship system: e global
etymon kaka. Part II: An anthropological study. Mother Tongue, 7, 245–258.
Matthey de l’Etang, A., & Bancel, P.J. (2005). e global distribution of (p)apa and (t)ata and
their original meaning. Mother Tongue, 9, 133–169.
Matthey de l’Etang, A., & Bancel, P.J. (2008). e age of mama and papa. In J.D. Bengtson (Ed.),
In hot pursuit of language in prehistory. Papers in the four elds of anthropology in honor of
Harold C. Fleming (pp. 417–438). Amsterdam: John Benjamins.
Pierre J. Bancel & Alain Matthey de l’Etang
Matthey de l’Etang, A., & Bancel, P.J. (In preparation). e Proto-Amerind kinship system. Com-
munication to be presented at the American Anthropological Association 111thAnnual
Meeting, to be held in San Francisco, CA, November 14–18, 2012.
Matthey de l’Etang, A., Bancel, P.J., & Ruhlen, M. (2010). Back to Proto-Sapiens (Part 1). e
inherited kinship terms papa, mama and kaka. In D. Jones & B. Milicic (Eds.), Kinship,
language, and prehistory. Per Hage and the renaissance in kinship studies (pp.29–37). Salt
Lake City, UT: University of Utah Press.
McDougall, I., Brown, F.H., & Fleagle, J.G. (2005). Stratigraphic placement and age of modern
humans from Kibish, Ethiopia. Nature, 433, 733–736.
Meyer, G. (1891). Etymologisches Wörterbuch der albanesischer Sprache. Strasbourg: Trübner.
〈http://www.archive.org/stream/etymologischesw00meyegoog〉 (16 June 2010).
Meyer-Lübke, W. (1911). Romanisches etymologisches Wörterbuch. Heidelberg: Carl Winter.
<http://ia311218.us.archive.org/2/items/romanischesetymo00meyeuo/romanischesety-
mo00meyeuo.pdf (15 April 2010).
Mumtaz, A. (1985). Baluchi glossary. A Baluchi-English glossary. Elementary level. Kensington,
MD: Dunwoody. 〈http://dsal.uchicago.edu/dictionaries/mumtaz/〉 (31 May 2010).
Murdock, G.P. (1959). Cross-language parallels in parental kin terms. Anthropological Linguistics,
1, 1–5.
Nebrija, E.A. de. (1492). Lexico o diccionario latino-español. Online facsimile. 〈http://www.cer-
vantesvirtual.com/servlet/SirveObras/07032774389636239647857/ima0192.htm〉 (12 June
2010).
Nikolayev, S. (2007). Indo-European etymology. Online etymological database, Tower of Babel
Project. 〈http://starling.rinet.ru/cgi-bin/main.cgi?ags=eygtnnl〉 (26 May 2010).
Nişanyan, S. (2001). Sözlerin Soyağacı. Çağdaş Türkçenin Etimolojik Sözlüğü. Istanbul: Everest.
〈http://www.nisanyansozluk.com/〉 (12 June 2010).
O’Connell, J.F., & Allen, J. (2004). Dating the colonization of Sahul (Pleistocene Australia-New
Guinea): A review of recent research. Journal of Archaeological Science 31, 835–853.
Oller, D.K. (1980). e emergence of the sounds of speech in infancy. In G.H.Yeni-Komshian,
J.F. Kavanagh, & C.A. Ferguson (Eds.), Child phonology: Vol. 1, Production (pp. 93–112).
New York, NY: Academic Press.
Pagel, M. (2000). Maximum likelihood models for glottochronology and for reconstructing
linguistic phylogenies. In C. Renfrew, A. McMahon, & L. Trask (Eds.), Time depth in his-
torical linguistics (pp. 189–207). Cambridge, UK: McDonald Institute for Archaeological
Research.
Pehrson, R.N. (1966). e social organization of the Marri Baluch. New York, NY: Wenner-Gren
Foundation for Anthropological Research.
Pokorný, J. (1959). Indogermanisches etymologisches Wörterbuch. Bern: Francke.
Raverty, H.G. (1867). A dictionary of the Puk’hto, Pus’hto, or language of the Afghans. With remarks
on the originality of the language, and its anity to other oriental tongues (2nd ed.). London:
Williams and Norgate. 〈http://dsal.uchicago.edu/dictionaries/raverty/〉 (31May 2010).
Ringe, D.A. (2002). Review of Greenberg (2000). Journal of Linguistics, 38, 415–420.
Rosenfelder, M. No date. Numbers in 5020 Languages. In e Metaverse. 〈http://www.zompist.
com/numbers.htm〉 (21 March 2011).
Ruhlen, M. (1994a). e origin of language. Tracing the evolution of the mother tongue. New York,
NY: John Wiley.
Ruhlen, M. (1994b). On the origin of languages. Studies in linguistic taxonomy. Stanford, CA:
Stanford University Press.
Brave new words
Rybatzki, V. (2006). Die Personennamen und Titel der mittelmongolischen Dokumente. Eine
lexikalische Untersuchung. Ph.D. dissertation. Helsinki: University of Helsinki. 〈https://
oa.doria./bitstream/handle/10024/828/dieperso.pdf?sequence=1〉 (26 May 2010).
Schulze, W. (2000). Northern Talysh. Munich: Lincom.
Schurmann, F. (1962). e Mongols of Afghanistan. An ethnography of the Moghôls and related
peoples of Afghanistan. e Hague: Mouton.
Singer, R., & Wymer, J. (1982). e Middle Stone Age at Klasies River Mouth in South Africa.
Chicago, IL: University of Chicago Press.
Sköld, H. (1936). Materialien zu den iranischen Pamirsprachen. Lund: Gleerup.
Starostin, S. (2006). Chinese characters. Online etymological database, Tower of Babel Project.
〈http://starling.rinet.ru/cgi-bin/main.cgi?ags=eygtnnl〉 (26 May 2010).
Stern, C., & Stern, W. (1907). Die Kindersprache: Eine psychologische und sprachtheoretische
Untersuchung. Leipzig: Barth.
Strand, R. (1997–2012). Kinship systems of the Hindu-Kush. In R. Strand, Nuristân. Hidden
Land of the Hindu-Kush. 〈http://nuristan.info/〉 (3 May 2010).
Trask, L. (2004). Where do mama/papa words come from? University of Sussex Working Papers
in Linguistics and English Language. 〈http://www.sussex.ac.uk/english/research/projects/
linguisticspapers〉 (21 January 2012).
Turner, R.L. (1962–1966). A comparative dictionary of Indo-Aryan Languages. London: Oxford
University Press. 〈http://dsal.uchicago.edu/dictionaries/soas/〉 (2 May 2010).
Vocabularium Cornicum. 12th century anonymous manuscript. 〈http://www.carlaz.com/
cornish/voccorn.txt〉 (12 April 2010).
Wang, W. S.-Y. (2004). Chinese dialects. Online etymological database, Tower of Babel Project.
〈http://starling.rinet.ru/cgi-bin/main.cgi?ags=eygtnnl〉 (16 May 2010).
We rn e r, B . ( 20 0 9 ). Terminologie der erweiterten Familienstruktur der Zaza (Region Çermik-Gerger-
Siverek). 〈http://www.zazaki.de/zazaki/Kinshipdiagram_deutsch.pdf〉 (31May 2010).
Westermarck, E. (1891). e history of human marriage. London: Macmillan.
White, T.D., Asfaw, B., DeGusta, D., Gilbert, H., Richards, G.D., Suwa, G., et al. (2003). Pleisto-
cene Homo sapiens from Middle Awash, Ethiopia. Nature, 423, 742–747.
Appendices: Comparative data
Appendix A. e Proto-Indo-European root *ma- ~ *mama- ‘mother’
[or, rather, ‘mother, mom’]
e reference of data not drawn from Nikolayev (2007) is given when it is relatively
dicult to access (i.e. essentially for the Indic, Nuristani, and Iranian groups); addi-
tional data from European language groups have been drawn from standard dictionar-
ies, oen accessible on the Internet.
Indic: Proto-Indic *mā ‘mother’: Pali māmikā ‘mother’; Prakrit māu ‘mother’; Germany
Gypsy mama ‘mother’; Romania Gypsy māmi ‘grandmother’; Bashkarik mêm
‘mother’s mother’, mâm ‘mother’s father’; Phalula mḗmi ‘mother’s mother’, mṓmo
‘mother’s father’; Domaki māma ‘mother’; Tirahi mā ‘mother’; Shina (Gilgiti
dial.) mā ‘mother’; Shina (Kohistani, Palesi) mā ‘mother’; Shina (Guresi) māh
Pierre J. Bancel & Alain Matthey de l’Etang
‘mother’; Sindhi māu ‘mother’; Lahnda mā ‘mother’; Lahnda (Awankari dial.) mā
‘mother’; Punjabi mā" ~ mā"u ~ māī ~ māmmī ‘mother’; West Pahari (Curahi dial.)
mā ‘mother’; Kotgarhi mā ‘mother’, māi ‘mother, goddess Durga’; Kumauni mā
‘mother, mother-in-law’; Nepali māu ‘female animal having given birth’; Assamese
mā ~ māu ‘mother’, māi ‘mother, mother’s brother’s wife’; Bengali mā ‘mother’,
māi ‘breast’; Oriya māā ~ mā ‘mother’, māi ‘woman’; Maithili māī ‘mother’; Bhoj-
puri māī ‘mother’; Awadi (Lakhimpuri dial.) māī ‘mother’; Hindi mā ~ māī ~ mā"
‘mother’; Old Marwari mā ‘mother’; Gujarati mā ~ māi ‘mother’; Marathi mā ~
māī ‘mother’, mā"ī ‘mother-in-law’ (Turner 1962–1966, etym. 10016 & 10058).
Iranian: Ossetic mamæ ‘mom’ (Abaev 1970); Yaghnobi momo ‘grandmother’ (Bird 2006);
Wakhi mum ‘grandmother’ (Grierson 1920); Persian mām ‘mom’, māmā ‘midwife’
(Hayyim 1934–1936); Zaza ma ‘mother’ (Werner 2009).
Armenian: mam ‘grandmother’.
Hellenic: Classical Greek mā! g ā! ‘Earth Mother!’; (Homeric) mâia ‘address to an old woman’;
(Attic) mâia ‘mom, wet nurse, midwife’; mámmē ‘mom, granny’; (Doric) mâia
‘grannny’; Standard Modern Greek mama ‘mom’, mammi ‘granny’.
Slavic: Proto-Slavic *mama ‘mom’: Belorussian mama ‘mom’; Russian mama ‘mom’;
Ukrainian mama ‘mom’; Bulgarian mama ‘mom’; Serbo-Croatian mama ‘mom’;
Slovene mama ‘mom’; Czech mama ‘mom’; Slovak mama ‘mom’; Polish mama
‘mom’; Upper Sorbian mama ‘mom’; Lower Sorbian mama ‘mom’.
Baltic: Proto-Baltic *mama ‘mom’; Lithuanian mama, (dial.) mōmà ‘mom’; Latvian mãma
‘mom’.
Germanic: Standard German Mama ‘mom’, Oma ‘granny’; Alemannic Mamme ‘mom’; Alsa-
tian Màmma ‘mom’; Low German Mame ~ Mamme ~ Mammä ‘mom’; Dutch ma ~
mam ~ mama ‘mom’, oma ‘granny’; Danish mama ‘mom’; Swedish mamma ‘mom’;
Norwegian mamma ‘mom’; Faeroese mamma ‘mom’; Icelandic mamma ‘mom’.
Italic: Latin Maia ‘Great Goddess = Earth, associated with the cult of Vulcan, and mother
of Mercury’, Maius ‘month of May’, mamma ‘mommy, mother, wet nurse’; Roma-
nian mámă ‘mother, mom’; Italian mamma ‘mom’; Sursilvan mumma ‘mom’;
Sutsilvan moma ‘mom’; Surmiran mamma ‘mom’; Putér mamma ‘mom’; Vallader
mamma ‘mom’; Friulian mame ‘mom’; French maman ‘mom’, mémé ~ mamie
‘granny’; Occitan mamà ‘mom’, mameta ‘granny’; Catalan mamà ‘mom’; Spanish
mamá ‘mom’; Portuguese mamãe ‘mom’.
Celtic: Proto-Celtic *mammā: Old Irish mam ‘mother’; Welsh mam ‘mother’; Breton mam
‘mother’; Cornish mam ‘mother’; Proto-Celtic *mammiā: Old Irish muimme ‘foster
mother’.
Albanian: Tosk m"m" ‘mother’; Gheg mam" ‘mother’.
Appendix B. e Proto-Indo-European root *pa ~ *papa ‘father, dad’
Anatolian: Palaic pāpa ‘father’.
Indic: 1. Proto-Indic *bappa: Prakrit bappa ‘father’; Armenia Gypsy bap ‘father’; Dam-
eli bàp ‘father, grandfather’; Gawar-Bati bāp ‘father’; Torwali bāp ‘father’; Lahnda
bāpū ‘grandfather’; Punjabi bāp, bāpū ‘father’; Nepali bāp ‘father’; Assamese bāp
‘father’, bāpā ‘term of address to a father or of aection to a young man’, bāpu ‘term
of address to a learned Brahman’; Bengali bāp ‘father’, bāpu ‘father, child’; Oriya
bāpa ‘father’, bapā ‘term of endearment to younger persons’, bāpu ‘term of address
Brave new words
to a father or to a young person’, (Puri dial.) bāpā ‘father’s father’; Maithili bāp,
bappā ‘father’; Awadi (Lakhimpuri dial.) bāp ‘father’; Hindi bāp ‘father’; Guja-
rati bāp ‘father’; Marathi bāp ‘father’; Sinhalese bapa ‘father’; West Pahari (Koci
dial.) bāp ‘father’, (Kiuthali) bapu (used by Rajputs), bāpū; Maldivian (upper class)
bappa, (lower class) bafā ‘father’. 2. Proto-Indic *babba: Domaki baba ‘father,
father’s brother’ (pl. piāra < pitŕ); Pashai (Areti dial.) bāba ‘father’; Shumashti
bā́bā; Bashkarik bab ‘father’, bobə ‘father’s brother’; Savi bāb, bābu ‘father’; Phal-
ula bā́bu ‘father’, bābá ‘father’s brother’; Shina (Gilgiti dial.) bābu ‘father’, (Palesi)
bubā; Kashmiri bab ‘father, grand-father’, bāb ‘father’, (Rambani dial.) babb ‘father’,
(Poguli) baub ‘father’, (Dodi) babbō ‘father’; Punjabi bābbā ‘father, grandfather’,
bābū ‘term of respect’, (Kangra dial.) babb ‘father’; West Pahari (Bhadrawahi dial.)
bābō ‘father’, (Bhalesi) bāb ‘father’, (Curahi) bābb ‘father’, (Cameali) babb ‘father’,
(Khashali) babb ‘father’ (voc. bāvā); Kumauni bābu ‘father’, babā ‘aectionate term
for father or child’; Nepali bābu ‘father’, bābai ‘term of address to child’, babuwā
‘father, (Tarai dial.) aectionate term for son’; Bengali bābā ‘father, baby’, bābu
‘gentleman’; Oriya bābā ‘father’, babā ‘father’s elder brother’, bābu ‘gentleman’,
babuā ‘term of endearment to juniors’; Maithili bābā ‘father’, bābu ‘title of respect’;
Hindi bābu ‘father’, babuwā ‘child’; Gujarati bābū ‘term of respect’; Marathi bābū
‘term of respect’; Marathi bābdā ‘term of endearment to a child’; West Pahari (Koci
dial.) bāb ‘father’, (Kiuthali) babu ‘father’ (used by Rajputs), bābū ‘father’. (Turner
1962–1966: etym. 9209)
Nuristani: Kâta-vari (Ktivi dial.) vov ‘grandfather’; Kâmv’iri vov ‘grandfather’; Supu-vari vâ
‘grandfather’; Sanu-vîri bâba ‘elder brother’; Usüt-vare vâv ‘grandfather’, bab ‘elder
brother’; Vä-alâ bâba ‘elder brother’; Ameš-alâ bâba ‘elder brother’; Nišei-alâ bâba
‘elder brother’. (Strand 1997–2008)
Iranian: Khwarezmian papa, bâb ‘father’ (Rybatzki 2006); Sogdian bâbay ‘father’ (Rybatzki
2006); Yaghnobi bobo ‘grandfather’ (Bird 2006); Bactrian babu ‘masc. personal
name’ (Rybatzki 2006); Pashto bâbû ‘dad, address term to an elder’, bābā ‘grand-
father’ (Kabir & Akbar 1999; Schurmann 1962); Wakhi pūp ‘grandfather’ (Gri-
erson 1920); Sanglechi bobo ‘father’s father’ (Rybatzki 2006); Ishkashmi bōbō
‘grandfather’ (Grierson 1920); Shughni bu0b ‘grandfather’ (Sköld 1936); Bajui bo0b
‘grandfather’ (Sköld 1936); Sahdara bo0b ‘grandfather’ (Sköld 1936); Bartangi bo0b
‘grandfather’ (Sköld 1936); Yazghulami ba0b ‘grandfather’ (Sköld 1936); Parachi
bâw ‘father’, bâbâ ‘grandfather’ (Rybatzki 2006); Pahlavi bâbâ ‘rst part of masc.
name’ (Rybatzki 2006); Farsi bâbâ ‘father, grandfather’ (Rybatzki 2006); Basseri
Farsi ba0 ‘father’, bābō grandfather’ (Rybatzki 2006); Dari bâbâ ‘grandfather, father,
dad’ (Rybatzki 2006); Tajik baba ~ bawa ~ baab ‘father’, bâbâ ‘ancestor’ (Schur-
mann 1962); Baluchi bâbâ ‘elder man’ (Rybatzki 2006); Marri Baluch baba ‘father,
grandfather, grandfather’s brother’ (Pehrson 1966); Hazara bâbá ‘father’ (Schur-
mann 1962); Kurdish bav ‘father’, bavo ‘dad’, bapir ‘grandfather’ (Rybatzki 2006);
Zaza bao ‘dad (vocat.)’ (Keskin no date).
Armenian: pap ‘grandad’.
Hellenic: Classical Greek pappa ‘dad’, pappos ‘grandfather, forebear, ancestor’; Modern
Pontic Greek papa ‘dad’ (Fauvin & Nikaki, personal communication); Standard
Modern Greek baba ‘dad’ (borrowed from Turkish, see Chantraine 1968), pappoús
‘grandfather’.
Baltic: Latvian paps ‘dad’.
Pierre J. Bancel & Alain Matthey de l’Etang
Germanic: Gothic papa ‘dad’; Modern High German Papa ‘dad’, Opa ‘grandad’; Alsatian Pàpa
‘dad’; Alemannic Pappe ‘dad’; Rhine Franconian Pàppe ~ Bàbbe ‘dad’; Bavarian
Babba ‘dad’; Dutch pa ~ papa ~ pappa ‘dad’, opa ‘grandad’; English papa; Danish
papa ‘dad’; Swedish pappa ‘dad’; Norwegian pappa ‘dad’; Faeroese pápi ‘dad’; Ice-
landic pabbi ‘dad’.
Italic: Latin pappa ‘dad’, pappus ‘grandfather, ancestor’; French papa ‘dad’, pépé ~ papy
‘grandad’; Sursilvan bab ‘father’; Sutsilvan bab ‘father’; Surmiran bab ‘father’;
Putér bap ‘father’; Vallader bap ‘father’; Friulian pai ‘dad’; Italian papà ~ babbo
‘dad’; Occitan papà ‘dad’, papet ‘grandad’; Catalan papà ‘dad’; Spanish papá ‘dad’;
Portuguese pai ~ papá ~ papai ‘dad’.
Albanian: baba ‘dad’ (borrowed from Turkish, Meyer 1891).
Appendix C. e Proto-Indo-European root *tat- ~ *tet- ‘father’
[or, rather, *tata ‘dad, father’]
Anatolian: Hieroglyphic Luwian tati(a)- ‘father’; Luwian tati(ja)- ‘father’; Lycian tedi ‘father’.
Indic: Sanskrit tātá ‘(vocative) aectionate address to junior’ (Śatapatha Brāhmana), ‘idem
to senior’ (Mahābhārata), ‘father’ (ibid.), tatá ‘father’ (Rig Veda); Pali tāta ‘term of
respectful or aectionate address to an elder or younger’; Prakrit tāa ‘father, son’;
Germany Gypsy tatta ‘father’; Pašai (Darrai-i Nūr and Wegali dial.) tatī́ ‘father’;
Khowar tat; Old Gujarati tāya m. (Turner 1962–1966: etym. 5754). Proto-Indic
*dādda ‘father or other elderly relative’: Germany Gypsy dād ‘father’; Domaki dādo
‘grandfather’; Dameli dádi ‘father’; Pašai (Laurowani dial.) dadā́ ‘elder brother’,
(Gulbahari) dādā ‘father’, (Kurangali) dādo ‘father’s brother’; Kalasha dā́da ‘father’;
Bashkarik dād ‘grandfather’, dēd ‘grandmother’; Phalula dōdo ‘father’s father’,
dēdi ‘father’s mother’; Shina dādo ‘grandfather’, dādi ‘grandmother’; Sindhi ḍāḍō
‘father’s father’, ḍāḍī ‘father’s mother’, (Kacchi dial.) ḍāḍo ‘grandfather’; Lahnda
ḍāḍā ‘father’s father’, 0ḍī ‘father’s mother’, dādā m., 0dī f.; Punjabi dāddā, dā0 m.,
dāddī, dī0 f.; Western Pahari (Bhalesi dial.) dādo m., (Kotgarhi) dād ‘father’s father,
elder brother’, daddi ‘father’s mother’, (Kiuthali) dādā ‘grandfather’; Kumauni dādā
‘grandfather, elder brother’, dādī ‘grandmother, elder sister’, dā ‘address to an elder
brother’; Nepali dādā ‘old servant’, dājyu, dāi (contaminated by bhāi < bhrā́tr0
[Proto-Indic form of PIE *bhratər ‘brother’, PJB & AME]) ‘elder brother’; Assamese
dādā ‘elder brother’; Bengali dādā ‘grandfather, elder brother’, dādi ‘grandmother’;
Oriya dādā ‘grandfather, father’s brother, elder brother’; Maithili dādā ‘grandfa-
ther’; Hindi dādā ‘father’s father, elder brother’, dādī ‘father’s mother’; Gujarati dād%
‘father’s father’, dādīf.; Marathi dādā ‘elder brother’, dādī ‘respectful term for an old
woman’. (Turner 1962–1966: etym. 6261)
Nuristani: Kâta-vari (Ktivi dial.) to ‘father’, -to ‘father’s (brother)’; Kâmv’iri tot ‘father’, -tot
‘father’s (brother)’; Vä-alâ tâta ‘father’, -ta ‘father’s (brother)’, el-ta ‘grandfather’
(cf. ei ‘mother’, el-ei ‘grandmother’); Ameš-alâ tâta ‘father’, -tâta ‘father’s (brother)’,
ga4-ta ‘grandfather’ (cf. ga4-ei ‘grandmother’); Nišei-alâ tâti ‘father’, -tâti ‘father’s
(brother)’ (Strand 1997–2008).
Iranian: Old Avestan tā ‘father’ (Yasna 47.3); Yaghnobi dodo ‘father’ (Bird 2006); Shughni
tat ‘father’ (Mancino no date); Roshani taat ‘father’, tatek ‘grandfather’; Ishkashmi
tot ~ tāt ‘father’ (Grierson 1920); Wakhi tat ‘father’ (Grierson 1920); Zebaki tā0
Brave new words
t ~ tā0’ father’ (Grierson 1920), Pashto dādā ‘a term of endearment to a father or
elder brother (East), also elder sister (West)’ (Raverty 1867); Zaza ded ‘father’s
brother’, dedo ‘idem (voc.)’ (Werner 2009); Talysh dada ‘father’ (Schulze 2000);
Baluchi dada ‘father’s father’ (Mumtaz 1985).
Greek: Classical Greek (Myrin.) tatā (voc.) ‘daddy’, Homeric tétta.
Slavic: Proto-Slavic *tata ‘father, dad’: Pskov, Arkhangelsk, Eastern and Southern dialects
of Russian tata ‘father, dad’; Bulgarian tato ‘father, dad’; Serbo-Croatian táta ‘father,
dad’; Slovenian táta ‘father, dad’; Czech táta ‘father, dad’; Polish táta ‘father, dad’;
Lower Sorbian táta ‘father, dad’; Upper Sorbian táta ‘father, dad’.
Baltic: Proto-Baltic *tet-ia-, *tẽt-iã-: Lithuanian tẽti-s ‘father’, tẽtẽ ‘dad’; East Lithuanian
tẽtē ‘father’; Samogitian tìti-s, dial. táiti-s ‘father’; Latvian tēte, tētis ‘dad’.
Italic: Latin tăta ‘dad’; Old Castilian taita ‘dad’ (Nebrija 1492), Old Catalan taita ‘dad’;
Catalan (dialectal) tata ‘dad, brother’; Neapolitan tàta ‘dad’; Romanian tátă
‘father, dad’; Sursilvan tat ‘grandfather’; Sutsilvan tat ‘grandfather’; Surmiran tat
‘grandfather’.
Celtic: Old Cornish tat ‘father’ (Vocabularium Cornicum c.1250); Cornish tat ‘father’;
Middle Welsh tad ‘father’ (Charles-Edwards 2003); Welsh tad ‘father’, dada ‘dad’;
Middle Breton tat ‘father’ (Izard 1965); Breton tad ‘father’, tata ‘dad’; Old Irish data
‘foster father’ (Charles-Edwards 2003).
Albanian: tat" ‘father’.
Appendix D. e Proto-Dravidian root *appa ‘dad, father’
Tamil appan$, appu ‘father (term of endearment used to little children or inferiors)’, appacci
‘father’, appāttai ‘elder sister’, appi ‘mistress of house, elder sister’; Malayalam appan
‘father’, appu ‘aectionate appellation of boys’; Kannada appa ‘father (frequently added
to the proper names of men as a term of common respect; used endearingly to children
by elders)’, apa ‘father’, appu ‘aectionate appellation of boys’; Kodagu appë ‘father’; Tulu
appa, appè ‘ax of respect added to proper names of men’, appè ‘mother’, appa ‘a mode
of calling a mother’; Telugu appa ‘father, mother, elder sister (frequently added to names
of men as a term of common respect)’; Kolami appa ‘father’s sister’; Gondi āpōṛāl ‘father’,
maipō ‘my father’, mī-āpō ‘thy father’; Maria tape ‘father’; Konda tappe, (L.) tāpe ‘father’
(Voc. 1656); Koya Su. tappe ‘(his, her) father’; Konda aposi ‘father (with reference to third
person)’. (Burrow & Emeneau 1984, etym. 156)
Appendix E. e Proto-Turkic roots *ata ‘dad, father’, *apa ‘dad, father’,
and *ana ‘mom, mother’
1. Proto-Turkic *ata ‘father’: Old Uighur ata ‘father’; Sary-Yughur ata ‘father’; Nogai ata
‘father’; Oirat ada ‘father, ancestor’; Karakhanid ata ‘father’; Turkmen ata ‘father’s father’;
Azeri ata ‘father’; Balkar ata ‘father’; Tuvin a’da ‘father’; Middle Turkish ata ‘father’; Tatar
ata ‘father’; Kumyk ata ‘father’; Tofalar ada ‘father’; Uzbek %ta ‘father’; Kirghiz ata ‘father,
ancestor’; Karakalpak ata ‘ancestor’; Modern Turkish ata ‘ancestor’; Bashkir ata ‘father’;
Uighur ata ‘father, ancestor’; Urum ata ‘father’; Cuman ata ~ atta ‘father’; Kazakh ata
‘father’; Khakassian ada ‘father’; Karaim ata ‘ancestor’. (Dybo 2006)
Pierre J. Bancel & Alain Matthey de l’Etang
2. Proto-Turkic *apa ‘father’: Orkhon apa ‘ancestors’; Old Uighur apa ‘ancestors’; Salar aba ~
apa ‘father’; Bashkir (dial.) apa ‘father’; Sary-Yughur awa ‘father’; Khakassian aba ‘father’;
Karakhanid apa ‘father, ancestor’; Tatar (dial.) aba ‘father’; Tuvin ava ‘father’; Turkish aba
‘father’; Kirghiz aba ‘father’; Altai aba ‘father, bear’; Azeri (dial.) aba ‘father’; Balkar appa
~ aba ‘father’; Chuvash oba ‘bear’; Turkmen (dial.) aba ‘father’. (Dybo 2006)
3. Proto-Turkic *ana ‘mother, mom’: Old Uighur ana ‘mother’; Karakhanid ana ‘mother’;
Azeri ana ‘mother’; Dolgan ińe ‘mother’; Gagauz ana ‘mother’; Turkmen ana ‘mother’;
Tuvin ij&e ‘mother’; Karaim ana ‘mother’; Middle Turkish ana ‘mother’; Khakassian inä
‘mother’; Kirghiz ene ‘mother’; Karakalpak ana ‘mother’; Oirat ene ‘mother’; Kazakh ana
‘mother’; Salar ana ‘mother’; Uighur ana ‘mother’; Chuvash ańne ‘mother’; Bashkir inä
‘mother’; Sary-Yughur ana ‘mother’; Kumyk ana ‘mother’; Yakut ij"e ‘mother’; Balkar ana
‘mother’. (Dybo 2006)
Appendix F. e origin of words for ‘dad’, ‘father’, ‘mom’, and ‘mother’
in the Chinese family.
e two databases (Starostin 2006; Wang 2004) from which the following data have
been drawn dier in their respective transcriptions of oral stops; we have aligned them
according to Wang’s transcription. Superscript numbers following modern dialectal
forms in Wang’s data transcribe tonal contours.
1. ‘Dad’: Preclassic paʔ ‘father’; Classic pá; Western Han pá; Eastern Han pwá; Early Post-
classic pwó; Middle Postclassic pwó; Late Postclassic pwó; Middle Chinese pwấ (Starostin
2006). Modern forms: Beijing pa3; Jinan pa3; Xi’an pa3; Taiyuan pa3; Hankou pa12; Chengdu
pa12; Yangzhou pai 3; Suzhou pa11; Wenzhou pa11; Changsha pa4; Shuangfeng po11; Nan-
chang pak41; Meixian pa11; Guangzhou pa11; Xiamen pa32; Chaozhou pa11; Fuzhou pa11;
Shanghai pa1 (Wang 2004).
2. ‘Father’: Preclassic paʔ; Classic pá; Western Han pá; Eastern Han pwá; Early Postclassic
pwó; Middle Postclassic pwó; Late Postclassic pwó; Middle Chinese pǘ (Starostin 2006).
Modern forms: Beijing fu
3; Jinan fu
3; Xi’an fu
3; Taiyuan fu
3; Hankou fu
3; Chengdu fu
3;
Yangzhou fu3; Suzhou fu32; Wenzhou voy22; Changsha fu31; Shuangfeng 'əu32; Nanchang
f
u32; Meixian fu3; Guangzhou fu32; Xiamen hu32 (lit.), pe32; Chaozhou pe22; Fuzhou xu32;
Shanghai vu32; Zhongyuan yinyun fu2 (Wang 2004).
3. ‘Mom’: Preclassic mhāʔ; Classic mhā́; Western Han mhā́; Eastern Han mhā́; Early Postclassic
mhṓ; Middle Postclassic mhṓ; Late Postclassic mhṓ; Middle Chinese mó (Starostin 2006).
Modern forms: Beijing ma11; Jinan ma11; Xi’an ma11; Taiyuan ma1; Hankou ma11; Chengdu
ma11; Yangzhou ma11; Suzhou ma11; Wenzhou ma22; Changsha ma11; Shuangfeng mo11;
Nanchang ma11; Meixian: ma11; Guangzhou ma11; Xiamen ma11; Chaozhou ma11; Fuzhou
ma11; Shanghai ma1; Zhongyuan yinyun ma2 (Wang 2004).
4. ‘Mother’: Preclassic mə'ʔ; Classic mə'́; Western Han mə'́; Eastern Han mə'́; Early Postclassic
mṓ; Middle Postclassic mə'́w; Late Postclassic mə'́w; Middle Chinese m(́w ‘mother’ (Staros-
tin 2006). Modern forms: Beijing mu2; Jinan mu2; Xi’an mu2; Taiyuan mu2; Hankou mu2;
Chengdu mu2; Yangzhou mo2; Suzhou mo11; Wenzhou mo22; Changsha mo2; Shuangfeng
mû2; Nanchang mu2; Meixian mu11; Guangzhou mou22; Xiamen bu2 (lit.), bo2; Chaozhou
bo21; Fuzhou mu2; Shanghai mu3; Zhongyuan yinyun mu2 (Wang 2004).
Brave new words
Appendix G. e descent of Proto-Indo-European *deik’e- ‘to show,
to point’ and *dekm)- ‘ten’.
1. Proto-Indo-European *deik’e- ‘to show, to point’:
Anatolian: Hittite: tekk-ussai- ‘to show’.
Indic: Proto-Indic *diś ‘to point’:
(a) Sanskrit (Rig Veda) diś-áti ‘points out’, díś ‘direction, region’, (Mahābhārata)
díś-ā ‘direction’; Pali diś-áti ‘points out’, disā ‘id.’, (Vājasaneyi-Sam6 hitā) dēśá ‘point,
region, part’; (Rāmāyaṇa) dēśá ‘province, country’; Prakrit dis-aï ‘tells’, dis-ā ‘direc-
tion’, dēsa ‘part, country’; Old Gujarati dis-i ‘direction’; Old Awadhi dih ‘direction’,
desa ‘country’; Armenia Romany les ‘earth, world, life’; Palestine Romany dēs ‘place,
camping ground’; Kalasha (Rumbur dial.) dēš ‘country’, dēša ‘far, distant’; Phalula
dēš ~ dīš ‘village’; Pashai dēš ‘cultivated eld’; Torwali diš-ā ‘towards’; Kashmiri
(Kashtawari dial.) diś ‘country’; Shina dǐš ‘place’; Sindhi d$ehu ‘country’; Western
Pahari (Bhadrawahi dial.) dēś ‘village’; Kumauni des ‘country’; Nepali des ‘country,
plains of India’; Bengali des ‘country’; Oriya desa ‘country’; Maithili des ‘country’;
Assamese dih ‘means, direction’; Hindi dis-na ‘to show, to exhibit’, dis ‘direction,
side’, des ‘country’; Marwari des-ṛo ‘small country’; Gujarati des ‘country’; Marathi
des ‘country’; Sinhalese das-aya ‘direction’, desa ‘country’. (Turner 1962–1966:
etym. 6339, 6340, 6547)
(b) Sanskrit (Kauśikasūtra) diṣ-ṭi ‘a measure of length’; Shina (Gilgiti dial.) di-ṭ
‘span’, (Jijelut dial.) dīṣ ‘span’; Dameli diṣ-ṭ ‘span’; Khowar diṣ-ṭ ‘handspan’; Kalasha
jiiṣ-ṭ ‘handspan’; Phalula diṣ-ṭ ‘span’. (Turner 1962–1966: etym. 6343)
Nuristani: Ashkun dešī’ ‘village’; Kalasha-ala (Waigali) deš ‘village’. (Turner 1962–1966: etym.
6340)
Iranian: (a) Old Avestan ā-diš-ti- ‘direction’; Avestan daēs ‘to show’, dax-šta- ‘sign, revela-
tion’; Khotanese dīs – ‘to confess’; Sogdian p-ð’ys ‘to show’; Parthian ‘dyš-g ‘sign’;
‘b-dys- ‘to show’; Ossetic dis ~ des ‘amazement’; æv-dīs-yn ~ æv-des-un ‘to show’.
(Lubotsky no date).
(b) Avestan diš-ti ‘a measure of length’; Khotanese di-ṭhi ‘a measure of length’.
Greek: (a) Class. Greek deík-numi ‘to show’; Cretan dik-nuti ‘to show’; Mod. Greek ðeíx-
no ‘to show’.
(b) Classical Greek (?) dak-tulos ‘nger’; Modern Greek (?) ðak-tilo ‘nger’.
Germanic: Proto-Germanic (a) *ga-tīhan ‘to announce, tell’, *taik-n ‘token’, *taik-njan ‘to
show, to manifest’; (b) *taih-wō ‘toe’:
(a) Old Norse tjā ‘to show’; Old High German zeig-ōn ‘to show’, zīh-an ‘to accuse’,
zeihh-an ‘sign’, ziht ‘accusation’; Old Franconian teik-in ‘sign’; Old Frisian tīg-ia
‘to show’, tēk-an ‘sign’; Old English tē-on ‘to show’, tä'c-an ‘to teach’, tä'c-en ‘sign’;
Middle High German zeig-en ‘to show’, zīh-en ‘to accuse’, zeich-en ‘sign, example’;
Middle Low German tīe-n (participe tīg-en) ‘to show’, tēk-en ‘sign’; Middle Dutch
tie-n ‘to show’, tēk-en ~ teik-en ‘sign’, tiht ‘accusation’; Icelandic tīg-n ~ teik-na ‘give
a sign’, Faeroese tek-na ‘to show’, tek-n ‘sign’; Norwegian te ‘to show’, teik-n ‘sign’;
Swedish te ‘to show’, teck-en ‘sign’; Danish te ‘to show’, teg-n ‘sign’; English teach,
tok-en; Dutch aan-tijg-en ~ op-tijg-en ~ be-tijg-en ‘to show’, tek-en ‘sign’, dial.
teiken ‘sign’; German zeig-en ‘to show’, Zeich-en ‘sign’, Alsatian zaig-a ‘to show’,
Zaich-a ‘sign’.
Pierre J. Bancel & Alain Matthey de l’Etang
(b) Old Norse tā ‘toe’; Old High German zēh-a ‘toe’; Old English tāh-e ‘toe’; Mid-
dle Low German tēwe ‘toe’; Icelandic tá ‘toe’; Faeroese tá ‘toe’; Norwegian tå ‘toe’;
Swedish tå ‘toe’; Danish tå ‘toe’; English toe; Dutch teen ‘toe’; German Zehe ‘toe’;
Alsatian Zecha ‘toe’.
Baltic: Proto-Baltic *teîg- ‘to tell’; Old Lithuanian tieg ‘he said’; Lithuanian téig-ti ‘to say,
tell, claim’.
Italic: (a) Oscan deík-um ‘to say’; Umbrian tik-amne ‘dedicace’; Latin dic-ere ‘to say’, dic-
tiō ‘discourse’, ju-dex ‘judge’ (telling ju-s, the law); French di-re ‘to say’, in-diqu-er
‘to show’; Occitan dís-er ~ digu-er ‘to say’; Catalan di-r ‘to say’; Aragonese dí ‘to
say’; Spanish dec-ir ‘to say’; Portuguese diz-er ‘to say’; Sursilvan di-r ‘to say’; Sutsil-
van gi-r [džir] ‘to say’; Surmiran dei-r ‘to say’; Putér di-r ‘to say’; Vallader di-r ‘to
say’; Friulian dî ‘to say’; Italian di-re ‘to say’; Romanian zic-e ‘to say’.
(b) Latin in-dex ‘indicative, index nger’, dig-itus ‘nger’; French doig-t ‘nger’;
Occitan de-t ‘nger’, en-dèi-s ‘index nger’; Catalan di-t ‘nger’; Spanish de-do ‘n-
ger’; Portuguese de-do ‘nger’; Sursilvan de-t ‘nger’; Sutsilvan de-t ‘nger’; Surmi-
ran de-t ‘nger’; Putér dau-nt ‘nger’; Vallader dai-nt ‘nger’; Friulian dê-t ‘nger’;
Italian di-to ‘nger’; Romanian deg-et ‘nger’.
2. Proto-Indo-European *dekm)- ‘ten’ (all reexes below also mean ‘ten’):
Indic: Proto-Indic *dasan; Vedic dáça; Prakrit dasa ~ daha; Pali dasa; Aśokan daśa ~
dasa; Apabhram6śa dasa ~ daha; European Romany deš; Armenia Romany las;
Palestine Romany das; Gondwani dhamak; Dameli daš; Domaki dai; Tirahi dā;
Poguli dāh; Rambani das; Kohistani daš; Pashai dāya; Shumashti däs; Ningalami
das; Wotapuri daš; Gawarbati d%š; Kalasha daš; Khowar jioš; Bashkarik daš; Torwali
daš; Kandia daš; Maiyā daš; Savi daš; Phalula dāš; Shina dái; Kashmiri dah; Ram-
bani das; Poguli dāh; Dodi dāś; Sindhi d$aha; Khatri dô; Kacchi d$au; Lahnda dāh;
Khetrani dā; Awankari dā; Punjabi das; Siraiki dah; Western Pahari daś; Kotgarhi
d́%ś; Garhwali das; Kumauni das; Nepali das; Assamese dah; Mayang dos; Bengali
das; Oriya dasa; Bihari das; Maithili das ~ dah; Magahi das; Bhojpuri das; Awadi
das; Lakhimpuri das; Hindi das; Bhili dōh ~ dah; Dogri das; Chattisgarhi das;
Khandeshi das; Braj das; Bundeli das; Urdu das; Rajasthani das; Malvi das; Magaji
das; Marwari das; Gujarati das; Marathi das ~ dahā; Konkani dhā; Sinhalese dasa-
ya ~ daha-ya; Maldivian diha. (Turner 1962–1966: etym. 6227; Rosenfelder no
date)
Nuristani: Kalasha-ala (Waigali) dōš; Wasi-weri lez; Kati duc; Kamviri d’uç; Ashkun dus.
(Turner 1962–1966: etym. 6227)
Iranian: Avestan dasa; Pahlavi dah; Khotanese dasau; Khwarezmian dhs; Turfanian dh;
Iron Ossetic dæs; Digor Ossetic dæs; Yaghnobi das; Pashto las; Wakhi ðas; Munji
dah; Ishkashmi da; Sanglechi das; Zebaki dos; Shughni ðīs; Yidgha los; Rushani
ðes; Yazgulami ðụs; Sarikoli ðes; Parachi ðōs; Ormuri das; Nayini de; Natanzi d’e;
Khunsari dēi; Gazi dē; Sivandi da; Vafsi dah; Semnani das; Gilaki da; Mazanderani
da; Talysh då; Harzani doh; Zaza des; Gorani da; Balochi dah; Southern Kurdish
da; Northern Kurdish da; Persian dah; Tajik daħ; Tati dæh; Chali dā; Farsi daśa;
Lari da; Luri dah; Kumzari da’hata.
Toch ari an: Tocharian A śäk; Tocharian B śak.
Armenian: Classical Armenian t’asn; Western Armenian tas.
Hellenic: Classical Attic Greek déka; Aeolic déko; Modern Greek ðéka; Tsakonian ðéka;
Cypriot ðéga; Pontic ðéka ~ réka.
Brave new words
Slavic: Proto-Slavic *des-ęt’; Old Church Slavonic des-ętĭ; Russian dés-jat’; Belorussian
dzés-jać; Ukrainian dés-jat’; Polish dzies-ięć; Kashubian dzes-ińc; Polabian dis-ąt;
Czech des-et; Slovak des-at’; Eastern Slovak dzeš-ec; Upper Sorbian dźes-ać; Lower
Sorbian źas-eś; Bulgarian dés-et; Serbo-Croatian dës-êt; Slovene des-et; Macedo-
nian des-et.
Baltic: Proto-Baltic *dēčim-t-; Old Prussian dessim-pts; Lithuanian dēšim-tis; Latvian
desm-its.
Germanic: Proto-Germanic *tíxun; Gothic taíhun; Old Norse tīu; Old Icelandic tīo; Old Swed-
ish tīo; Old Danish ti; Old High German zehan; Old Saxon tehan; Old Frisian
tian; Old Low Franconian tēn; Old English tēne; Middle Low German tein; Mid-
dle Dutch thien; Middle High German zehen; Icelandic tío; Norwegian tio ~ tie;
Swedish tio; Dalecarlian tiu; Faeroese tíggju; West Frisian tsien; Saterland Frisian
tjoon; Fohr North Frisian tjiin; Sylt North Frisian tiin; Helgoland North Frisian
tain; Dutch tien; Low Saxon tain; Westphalian Saxon tein; Crimean Gothic thiine;
English ten; German zehn; Bavarian zene; Swabian zaen; Cimbrian zègan; Rhine
Franconian zeen; Luxemburgish zéng; Swiss German zäh.
Italic: Latin decem; Old French dis; French dix; Walloon dijh; Jèrriais dgix; Picard dich;
Poitevin dis; Occitan dètz; North Occitan dié; Franco-Provençal dyî; Aragonese
deu, dech-igüeit ‘eighteen’; Catalan deu; Spanish diez; Ladino dies; Asturian diez;
Galician dez; Portuguese dez; Sursilvan diesch; Sutsilvan diesch; Vallader desch;
Friulian dîs; Ladin díesc; Piedmontese dés; Milanese dés; Genovese dexe; Venetian
diese; Corsican dece; Umbrian dèsce; Neapolitan riécë; Sicilian dècis; Italian dieci;
Sardinian deghe; Romanian zece; Arumanian date; Meglenian zeti.
Celtic: Gaulish decam; Old Irish deich; Irish deich; Scottish Gaelic deich; Manx jeih; Welsh
deg; Breton dek; Vannetais dek; Cornish dek.
Albanian: Standard Albanese dhjetë; Gheg ðet; Tosk zjétë.
