Skeptical Humanities

I take on board Kenneth G’s point that in many cases it is not clear that anybody actually gets hurt by a strange belief about language.  But I would argue, with Pacal, that there are cases where harm of various kinds can arise from such beliefs (reinforcement of racism, religious fundamentalism or extreme nationalism; reliance upon ineffective therapies; etc.).  And of course I would hold, as a ‘modernist’, that (other things being equal) it is better to adhere to (probably) true ideas than to false ones.

The comment about amateur attempts to find ‘cognates’ at the phoneme level which I mentioned last time was made by Ken, not by Kenneth G (sorry for the mix-up!).

To return to my theme:

There are several especially interesting sub-sets of ‘fringe’ historical linguistic claims of the type I’ve been discussing.  The first of these sub-sets involves the alleged mutual intelligibility of languages generally believed to have had no common ancestor in historic times and no significant pre-modern contact.  For example, Cyclone Covey and Ethel Stewart have claimed that Uighur (Turkestan) and Navajo and the Na-Dene languages generally (USA) are (or recently were) mutually intelligible, explaining this in terms of some Asian groups having migrated to the Americas much more recently than is normally supposed.  Other such claims involve Basque or Gaelic being mutually intelligible with languages of the Americas, Crimean Tatar being mutually intelligible with Latvian, etc., etc.  In a broadly similar vein, Gavin Menzies claims that Chinese is widely spoken or at least understood in unexpected places such as Peru, because of the alleged circumnavigation and exploration of the globe by 15^th-Century Chinese fleets.  But in no such case is there adequate evidence of genuine mutual intelligibility.

A related sub-set of cases is illustrated by Bruria Bergman’s claim that a temple chant used in Japan (with a trite and irrelevant meaning in Japanese, as is typical of such chants) is in fact in distorted Hebrew; in fact, it is much more easily, by way of a spoof, interpreted as, for instance, late Latin introduced by ‘Dark-Age’ Christian missionaries who are known to have been active in neighbouring parts of China (I myself was able to invent such a reading in twenty minutes).

There are two further special, overlapping groups of claims about: a) the deliberate or semi-deliberate conspiratorial concoction of languages or language data out of other known languages or reconstructed (or invented) languages (often by churches and other bodies with an alleged interest in deceiving humanity); and b) ancestor languages of a specific type involving very short words.  For cases of sub-type a), the relevant statistical considerations are much more difficult, since these considerations assume normal unplanned change.  These particular theories, although they are typically both implausible and indemonstrable, are thus almost immune to effective disproof along these lines.  However unsystematic and/or otherwise implausible a set of changes might be, it could occur if it was deliberately planned as part of a project of language concoction.

Examples of this type of claim include those of Polat Kaya (almost all words of almost all languages are really Turkish, deliberately corrupted and in this case often ‘anagrammatized’ so as to conceal their origin), Ior Bock (the source language is Finnish Swedish), Isaac Mozeson and others (Hebrew), Edo Nyland (Basque), P.N. Oak and his followers (Sanskrit), Michal Tsarion (Irish Gaelic, seen as a ‘secondary Ursprache’ arising after the fall of Atlantis), etc.  As before, the language identified as source is usually one favoured by the author, typically his own language or its ancestor.

By way of an additional feature (b), some cases of this special type, and a few of the otherwise ‘normal’ type as discussed above, involve the re-analysis of known linguistic forms (especially of the alleged ancestral forms) into sequences of monophonemic (single-phoneme), monosyllabic or other very short morphemes (meaningful word-parts) or words.   For example, some of those who assert that Hebrew is close to the Ursprache, and that the ultimate origins of Hebrew and Hebrew-derived words have been concealed, also propose monophonemic or other very short morphemes for early Hebrew.

Another such claim is that of John J. White, who uses the usual amateur philological and etymological methods to trace all languages back to an Ursprache called ‘Earth Mother Sacred Language’ (EMSL).  This alleged language had only very short morphemes.  White proposes fifteen basic morphemes; two of these are monophonemic (the vowels a and u), nine are monosyllabic (eight of these have the form Consonant-Vowel, the ninth is en) and the remaining four are disyllabic (Consonant-Vowel-Consonant-Vowel).

Each of EMSL’s very short morphemes has variant phonological forms (allomorphs).  One morpheme, de (basic sense ‘the’) has as many as nine variant initial consonants and various variant vowels.  The total number of ‘morphophonemic’ forms (word-shapes) is thus substantial; but all forms are allomorphs of the basic fifteen morphemes and derive their meanings from these.  For instance, za is a variant of de and therefore retains its sense (‘the’).  The supposed existence of so many very short and often widely differing forms with shared meanings obviously increases the freedom of the inventor enormously.  In particular, the acceptance of many variant vowels in each morpheme generates a theory in which the vowels, as White acknowledges, are often irrelevant.

In addition, in EMSL the ordering of these short morphemes is itself said not to be significant; a given sequence of morphemes will normally have the same meaning regardless of their linear order as spoken.  Thus both ty-re and ra-za – the same two morphemes, in opposite orders and also in variant forms – mean ‘the earth’.  This stipulation obviously increases the freedom of the inventor even further.

Some of the EMSL morphemes are transparently similar to short words with the same meaning in known languages; for example, ge (‘earth’) is very similar to the equivalent Ancient Greek form.  Others, such as ni (‘people’), are less obviously linked with known forms from familiar later languages and are more clearly the result of White’s very loose application of the comparative method.

White’s use of cross-linguistic data is typically naive.  For example, he deduces from the use of suffixed definite articles (the equivalents of ‘the’) in Albanian that this pattern may be of long standing in European languages generally (it is not), and he therefore frequently interprets final –de and its equivalents, in various languages, as a definite article; this includes the endings of English earth and the closely related German equivalent Erde.  He also interprets initial d(e)- in a similar way, for example treating de– in Latin deus (‘god’) and dea (‘goddess’) as an article – even though Latin has no definite article, meaning that these words, at least in their Latin guises, contain no material with that meaning.  In all such cases, White is of course implicitly rejecting known or very likely etymologies for the words in question.

More generally: it is possible for a few morphemes in a language to be very short, and indeed some may consist of only one phoneme.  These are usually grammatical morphemes/words rather than ‘lexical’ morphemes (normal dictionary words and word-roots); grammatical morphemes more generally are often especially short.  Most monophonemic morphemes involve vowel phonemes, as in English /ə/ = a (the indefinite article, as in a book), but this may also arise with certain types of consonantal phoneme, as in the case of Russian v, meaning ‘in’ or ‘to’.  (When speaking, linguists refer to the phone [ə] or the English phoneme /ə/ as schwa, from a Hebrew word.)

However, this pattern can apply to only a very few morphemes of any given language; and these monophonemic morphemes are still morphemes in precisely the same way as are other morphemes which consist of two or more phonemes.  Thus, English a contrasts grammatically and semantically with the diphonemic (two-phoneme) English definite article the, and is identified as a morpheme by the same methods of grammatical analysis.  (It itself also has a second, diphonemic allomorph: an before an initial vowel, as in an apple.)  Such morphemes are simply morphemes which happen to be monophonemic.  Where the phoneme /ə/ occurs in English words other than the indefinite article, for instance as the initial a– in around, it does not represent the indefinite article or its meaning.

And there are solid general linguistic reasons why the morphemes of a language cannot be predominantly monophonemic.  For instance, if most or all morphemes were monophonemic, the result would be a truly vast amount of homophony: different, unrelated morphemes/words with identical pronunciations.  Such a language would be unusable.  Even White’s system yields huge numbers of homophones.  (And even languages with predominantly monosyllabic morphemes and limited inventories of possible syllables – such as Chinese of various types – have had to develop special systems in order to have vocabularies of adequate sizes.)

Claims of this kind, like claims regarding the ‘concoction’ of languages, are highly implausible but very difficult to refute.  Very many longer words of known languages will contain each given sound, and it is not difficult to concoct accounts deriving the meanings of these words from those allegedly associated with each sound making up the longer, known word (especially where their order is said not to be significant, as in EMSL).  Where the etymologies do lend themselves to serious examination, the derivations are typically far-fetched and naturally in conflict with those generally accepted.

More next time on a particularly sensationalistic claim of this last type; then more special types of ‘fringe’ claim in this area.

Mark

As I said last time, I’ll now go into further detail on the issues discussed earlier and their significance for ‘fringe’ historical linguistic claims.  Some of the text presented here is based on sections of my forthcoming book on these matters.

The precise probability of accidental (unsystematic) similarity between words in different languages depends upon a number of factors:

a) the degrees of phonological and semantic similarity (phonetic form on the one hand and meaning on the other) required if words are to be seen as prima facie likely to be shared between languages

[Such decisions are often arbitrary to a degree; for example, scholars might disagree as to whether a word san meaning ‘scarlet’ in one language and a word zen meaning ‘orange’ in another language were similar enough in form and/or meaning to be regarded, prima facie, as probably shared (but note that the degree of phonological systematicity across the vocabularies, as discussed last time, will often be helpful in resolving such cases).]

b) the lengths of the words (for example, if two languages not known to be connected share a very short word-form such as [sa] with the same meaning, this could very well be accidental, whereas if they share a form such as [tolpesveblig], again with the same meaning, or with transparently related meanings, this is less likely)

[There are, of course, ‘borderline’ cases involving intermediate word-lengths; and where there is supporting non-linguistic evidence – especially where contact and borrowing are in question – some such cases may be fairly persuasive (even if systematicity cannot be demonstrated).  For example, three-syllable words of the form kumara or umara, meaning ‘sweet-potato’, occur on both sides of the South Pacific, and this vegetable, apparently of South American origin, has long been cultivated in Polynesia.  It is thus not unlikely that the sweet-potato diffused by means of unrecorded transoceanic contact; this would probably have involved intrepid (but non-literate ) Polynesian voyagers reaching what is now Peru/Chile and bringing the plant and borrowed names for it back to their homelands.}

c) the cross-linguistic frequencies of phonemes (speech-sounds) and phoneme-sequences (very widely-shared sounds such as [e], [s], etc. or common sound-sequences such as [til] or [po] are more likely to be shared by chance than sounds and sequences found in relatively few languages)

[I’m restricting myself here to international phonetic symbols (always given in [square brackets] by linguists) which correspond in form and approximate pronunciation with familiar Roman letters; I’ll introduce unfamiliar phonetic symbols only if necessary, and I’ll explain them when I first use them.]

d) the phonological and semantic systems of the relevant languages (these considerations are often rather technical where phonology is in question)

However, in any given case the precise probability of accidental (unsystematic) similarity is easily shown to be much greater than most non-linguists (including ‘fringe’ writers such as those discussed here) imagine.   There are millions of words and word-parts in the several thousand known languages; and there are only so many common sounds and sound combinations.  The calculable probability of pairs of superficially and unsystematically similar words in apparently unrelated languages having very similar or the same senses by chance is in fact much higher than non-linguists – including non-mainstream writers of the kind I’m discussing here – generally imagine.  Therefore, superficial phonetic similarity between isolated words and/or meaningful word-parts taken from different languages is in itself no evidence of cognatehood or of any genuine, non-accidental connection, even if there are many such words or if their meanings too are similar.  If the meanings are not really especially similar, or are merely alleged to be related as part of some writer’s theory, the case is even weaker.  There is vast scope for accidental similarity between the words of unconnected languages.

And indeed, as noted, it is phonological (and also grammatical) systems which are normally decisive in establishing genuine links, not superficially similar words per se.  But few ‘fringe’ writers know enough linguistics to deal adequately with phonology or grammar.  Indeed, the vast bulk of the ‘evidence’ associated with non-standard amateur claims involves vocabulary, which is replete with superficial similarities.  The analysis of vocabulary requires much less understanding of linguistic theory and methodology.

In fact, any major loosening of the standards of evidence for cognatehood, such as are required by claims such as those I’m discussing here, would have the consequence that very many alternative proposals (involving, for example, a whole range of different languages of origin for the same words) would be roughly equally plausible.  But these proposals all contradict each other; only one of them, if any, could be correct.  In that event, the reasonable conclusion would probably be that we could not say much at all about philology or older etymologies with any confidence.  Mainstream linguists would regard this conclusion as a last resort and as not warranted by the actual evidence.

In addition, etymologizing should (for obvious reasons) be based not upon well-known contemporary forms, as often occurs in the ‘fringe’ literature, but instead upon the forms of the relevant words in the oldest available versions of the languages in question – and upon the ranges of known cognate forms in the relevant language ‘families’.

Another issue here involves known or very well-grounded established etymologies for words, involving the often well-established histories of these language ‘families’.  Many of the novel etymological claims discussed here fly in the faces of known or very probable etymologies, which are often very well supported with historical and linguistic evidence.  For example, Michal Tsarion proposes an outlandish etymology for the English word bishop, ignoring the well-established Greek etymology; Gene Matlock advances equally bizarre Sanskrit etymologies for a wide range of words with known etymologies, for instance deriving the modern coining Australia (with its transparent Latin etymology: ‘southern [land]’) from Astralaya, supposedly meaning ‘land of missiles’.

Proposers of such alternative etymologies obviously need to argue that theirs are more plausible than the established ones.  But this is very seldom even attempted; readers are simply invited to accept the alternative etymologies, and the established ones are mentioned, if at all, only to be glibly dismissed.  This renders many of the etymologies offered seriously implausible or indeed impossible.

Other etymological claims deal mainly with the very remote past where the actual etymologies for words and word-parts are obscure and uncertain, or simply cannot be established, at least by current methods.  The point here is not that novel etymologies offered by non-mainstream writers are known to be wrong but that there is no particular reason to believe that they are correct.

In some cases multiple etymologies with different sources are posited for the very same word or morpheme.  For example, Don Smithana offers a double etymology for the –osis suffix in the Canadian lake-name Winnipegosis.  Etymological ‘blends’ of this kind do occur, but are predictably very rare; any such claim must be very strongly supported.

More next time on various special types of ‘fringe’ claim in this area.

Mark

Hi again, all!

Thanks very much for the appreciative and well-informed reactions to my first post!  I’ll start with some responses to the issues you’ve raised.

As has been noted, the theory of a recent Proto-World is in fact difficult to reconcile with various pieces of evidence, notably the modern archaeological evidence that modern humans had spread as far as remote Australia by 40,000 (or even 60,000) years BP, presumably already using languages related to other human languages (as are the contemporary Aboriginal languages).

Some enthusiastic commentators have asked for information about language ‘super-families’: larger, deep-time ‘families’, each giving rise to recognized later/smaller ‘families’.  One of you specifically mentioned Nostratic, which is a supposed deep-time common ancestor of Indo-European and some other ‘families’ including Semitic (Arabic, Hebrew etc.), Uralic (Finnish etc.) and Dravidian (Southern India); the exact composition of the Nostratic ‘super-family’ varies among its advocates.  Now down the track I may well have things to say about ‘super-families’ and about ‘Nostratic specifically; but many of the ‘Nostraticists’ and the advocates of other ‘super-families’ are themselves professional linguists (although they constitute minorities in accepting such theories), and much of this material is thus outside my remit here.  Controversial ideas are not by any means always ‘fringe’.  (But I still welcome comments and queries on these matters and will try to engage in discussions of them!)

On the other hand, there will definitely be more to say in this context about some specific languages which have attracted ‘fringe’ as well as mainstream attention, such as Etruscan (written in a known script but apparently ‘genetically’ isolated and still only marginally intelligible) and Sumerian (the earliest known written language and another which, while now largely understood, is again regarded as ‘genetically’ isolated) – also about the ‘cuneiform’ script used originally for Sumerian and later applied to unrelated Mesopotamian languages.  Comments on specific non-mainstream historical ‘linguists’ such as Jordan Maxwell and Michal Tsarion are also forthcoming! 

If readers’ patience with me holds out, I will also have something to say about ‘Inuit words for snow’ and related matters – but not under the present heading, since these issues are essentially non-historical in character.  If it appears after some time that I’m forgetting to get back to this, please nudge me!

OK, to resume:  As I said in Post 1, the language identified by ‘fringe’ writers as Proto-World may be a known (ancient) language or a language which has been reconstructed or invented by the author in question.   Where the language is a familiar one, it is usually an ancient language favoured [British spelling!] by the author, often the ancestor of her own language or even a known older stage of that language.  Biblical Hebrew, regarded by many as the Ursprache in pre-scientific times, remains popular in this respect, but there are many others, especially revered classical languages such as Sanskrit and Ancient Greek.  Languages of unknown or disputed ‘genetic’ affiliation, such as Hungarian, Sumerian and especially the genuinely mysterious Basque (another ‘genetically’ isolated language), are also commonly identified as ancestor languages.

The identification of a language dear to the writer as the Ursprache – or as otherwise especially important – does not of course itself show that the writer is biased, or even mistaken; but it excites reasonable suspicion as to unscholarly motives.

Other claims of this first sub-type do not involve an ultimate Ursprache but rather alleged more recent (but still ancient) ‘genetic’ links between specific cultures and languages which are normally deemed unconnected (one such writer is Michal Tsarion, who proposes a ‘secondary Ursprache’ which allegedly emerged after a catastrophe; see further later).

The second way in which unacknowledged links between languages are said to have arisen (see Post 1) involves ‘borrowing’ or ‘transfer’.  In cases of this kind, an individual word (or other linguistic feature) from one language is taken over by another language in a contact situation, as in genuine cases such as English restaurant, borrowed from French.  For example, members of one culture are said to have completed long voyages and to have arrived in the territory of the other, and the cultures and languages are supposed to have influenced each other.  (Claims of this kind clearly do not directly involve an Ursprache.) Some such cases involve the theories of Barry Fell, Thor Heyerdahl etc. about unrecognized early transoceanic contact.

In all these various types of case – even where language is not itself of especial interest to the writers in question – linguistic forms (spoken and/or written) are very commonly invoked, because they (with their meanings) appear prima facie much more specific and much more easily identifiable than most other cultural traits, and the probability of chance similarity seems much lower.  For instance, it is observed that the male name Madoc is found in Welsh and that the male name Modoc is found in Mandan (USA).  It is held that these two forms are so similar that they are very probably etymologically related; and on the basis of a limited number of individual pairs of this kind it is deduced that the Welsh and the Mandans had a common ancestor culture or experienced influential contact in remote times.  In this specific case, it is normally this latter which is at issue; the matter involves the American voyage of the medieval Welsh prince Madoc, an event not accepted as historical by mainstream scholarship.

In fact, in the absence of continuous textual evidence it can be established that words are cognates – that they descend from a common ancestor word or root in a common ancestor language, or that one of them is itself that common ancestor word – only if they display systematic correspondences in their phonology (the structural sound-units that make them up), repeated over large numbers of word-sets.  See for instance the systematic correspondences between synonymous (related) words, in the ‘subfamily’ of Indo-European languages derived from spoken Latin and known as ‘Romance’, in word sets such as Latin canis, French chien, Italian cane, Spanish can [later replaced by an unrelated word] (all meaning ‘dog’), Latin panis, French pain, Italian pane, Spanish pan (all meaning ‘bread’), Latin manus, French main, Italian mano, Spanish mano (all meaning ‘hand’), etc., etc.  In each word-set, Latin -an- has become –ain/-ien in French, –ane in Italian, –an in Spanish.  Such patterns exist because language change is very largely systematic and regular.

In ignorance of this, one might easily imagine, for instance, that Latin habere and German haben are cognates; after all, they are very similar (they differ only in respect of their grammatical endings), they both mean ‘to have’, and in this case we know independently that the languages themselves are ‘genetically’ related (that is, they have a common ancestor, the ancestor of all the Indo-European languages, although they are members of different ‘subfamilies’ of Indo-European).  But in fact these words are not cognate; they are unrelated, and their similarity is unsystematic and accidental.  German haben does have a Latin cognate, but this is capere (‘take’).  In fact, word-initial Latin c– systematically corresponds with German h-; other examples are canis/Hund (‘dog’), centum/hundert (‘hundred’), etc.

The reverse pattern also occurs; for example, English cow and beef, which now sound (and look, in spelling) very different, are in fact demonstrably systematically related and cognate; they do have a common deep-time (Indo-European) ancestor word.

Correspondences are less predictable where contact and borrowing are in question, partly because these involve interaction between two sound-systems and often individual words; but they are still fairly systematic, not merely haphazard, and any claim that a word has the phonological shape that it has because it has been borrowed from a given other language must be supported.

Next time I’ll go into further detail on these issues and their significance for ‘fringe’ historical linguistic claims.

Thanks again for all the interest!

Mark

Two key points here:

1) It is not certain that there ever was ONE single Proto-World; humanity may have developed language more than once.  It IS possible that all known human languages (whether still used or not) descend from one common ancestor, either because humanity did in fact develop language only once and the phenomenon diffused from that one starting-point, or because only one ultimate ancestor language, out of a number which once existed, has left any surviving descendant languages.  On the other hand, it is possible that the known languages descend from multiple ancestors. There are on present reckoning at least fifty ‘families’ of ‘genetically’ related languages (including some ‘genetically’ isolated single languages) which by definition cannot currently be shown to have ever had common ancestors; the best known such ‘family’ is Indo-European, which includes English, most other European languages ancient and modern, the North Indian languages descended from spoken Sanskrit, ancient and modern Persian, etc.  (Some linguists seek to establish deeper-time relationships between these ‘families’ by means of overtly statistical methods.)  We remain ignorant on this issue precisely because of Point 2).

2) As is generally agreed by historical linguists, on the evidence available, Proto-World or multiple ancestor languages must have been spoken so long ago (at least 70,000 years BP, probably more like 150,000) that (given the observed range of rates of linguistic change) it/they cannot possibly be reconstructed in any detail.  Using statistical methods such as those mentioned in Note 1, a few ‘maverick’ linguists HAVE argued that more recent dates for Proto-World should be accepted, and thus that Proto-World (along with other very ancient ancestor languages closely descended from it) CAN be reconstructed in part.  However, most of this work is now dated, and all of it is marginal to the mainstream at best; much of it involves methodology which, at least nowadays, is regarded by most linguists as too loose and approximate to be reliable.  Similar, often more extreme ideas have been developed by clearly non-mainstream authors.  These writers too consider that a single Ursprache/Proto-World (typically a language similar to their own ‘favourite’ ancient language; see later) existed relatively recently (within the last 20-30,000 years, or even more recently) and CAN be reconstructed in part.

The language identified as Proto-World may be a known (ancient) language or a language which has been reconstructed or invented by the author in question.  More on this next time!

Mark