A Language of 40,000 BC

2007 Justin B Rye

PREFACE – First Words

It seems to me that attempts to imagine the evolutionary origin of speech never make enough of an effort to account for the particular kind of language we've ended up with.  Alternatives may be far from obvious, but to demonstrate that they do exist, here's an illustrative “what‐if”.  It also happens to shed some light on a question first raised by Arthur Dent: why are cavemen so bad at Scrabble?

This page is a companion to the ones on Futurese, Alternese, and (more distantly) Europan; readers may also wish to compare and contrast my notes on SF Xenolinguistics and SF Chronophysics.  Or if you just want a run‐of‐the‐mill constructed language, there's a more detailed one over here.

Table of Contents:
PREFACE: First Words
INNATENESS: Early Talkers
PROTOLANGUAGE: Origin of Speeches
NEANDERTHALESE: Conlang of the Cave Bear
PHONOLOGY: The Voice of the Mousterians
GRAMMAR: Flint‐Axe Syntax
EXAMPLES: Towards Babel
HOLOCENESE: Scrabble for Survival

Unlike most of my language‐related pages, this one doesn't have much phonetically transcribed material, but what it does have is presented directly in Unicode IPA script, so apologies to anybody stuck using the sort of stone‐age web‐browser that can't cope with this.


There has never been any shortage of theories in the field of glossogeny (that is, the origins of language).  Back in the nineteenth century, there were so many that linguists had to declare the entire topic forbidden territory – it was a waste of effort to debate the theories when there was no evidence to work from!  That was a remarkably sensible decision, but the emergence of techniques like molecular genetics has changed things recently.

Unfortunately, the field also falls between two disciplines that have traditionally ignored one another.  There is still a tendency for theoretical linguists to see their subject‐matter as so super‐special that it couldn't possibly have been incrementally bodged into existence by anything as haphazard and biological as natural selection; it must be a pure chunk of spontaneously crystallised loveliness (to caricature Chomsky's apparent attitude).  And on the other side there has always been a tendency for biologists to overlook the stranger features of human language, confusing it with the sorts of communication that chimps are capable of (like learning to hammer the button and the button alternately), and taking it for granted that all they'd need is a few extra neurons and they'd automatically end up speaking English.

Another simple example of this failure of imagination: we evolved from bipedal great apes, world champions in hand/eye coordination with no special preadaptations for auditory processing or breath‐control – as far as we know, their vocalisations were essentially governed by involuntary reflex.  We have no reason to assume their development of more sophisticated modes of communication started with grunting; any sign‐language user could tell you the advantages gestural languages have over oral ones.  An early reliance on transparently iconic “body language” might help to explain the origin of intentional referential behaviour too (that is, verbally “pointing at” things).  Homo erectus showed some development of the brain regions used for language‐processing; and yet I frequently see it assumed that if its larynx wasn't like ours, then it must have been entirely dumb.

Meanwhile, some computational modellers have come up with simulations suggesting that it's easy for the right kinds of neural nets to self‐organise their way through the early steps in the development of verbal communication.  The problem with this is that in real life most species don't take those steps; on the contrary, if it was any rarer, there'd be nobody to comment on the fact!  This implies that the question is really: what were we doing with the “right kinds” of neural nets for processing combinatorial code‐systems built up in complex nested hierarchies?  Those things really don't grow on trees of the physical variety!  Things would be more explicable if only we could show that our early ancestors had some special problem (not shared by their primate cousins) to which a basic missing‐link protoparser was the natural solution.  There have been several suggestions…

The answer may well be that several of these functions were significant at different stages; after all, you don't get from a hooting ape‐creature to a Homer (or a Fred Flintstone) in a single leap.  There has to have been a consistent long‐term tendency for some individuals, or tribes, or species, to succeed because they had slightly fancier communicatory modalities… but the early stages are still utterly mysterious.  We have no idea to the nearest half‐million years when it was that the first hominid used a conventional symbolic name for a particular kind of entity, or coined an original word as opposed to an instinctive cry, or treated the order in which words were produced as significant.  We don't know which of these developments came first, or whether the communicative impulse behind them was sincere or manipulative, or how long it was before the innovators found people to talk to who were capable of understanding!

Well, I'm not going to pretend that I know better than any of the people doing research in this field; rather than attempt to present an Elaborated Handwaving Hypothesis for glossogeny, I'll concentrate on later stages, when language had developed into a state almost but not quite like what we're used to.

INNATENESS – Early Talkers

Just as babies seem to be born capable of recognising a human face, they also seem to have a natural urge to acquire a human language, and very reliably go through a developmental stage of doing so, regardless of whether or not anybody is trying to “teach” them one (unlike for instance kittens, which may overhear plenty of speech but will never think to try imitating it).  This is what linguists normally mean when they refer to “innateness”: not that you were born preprogrammed with knowledge of English, just that you were born with an aptitude for language.  Children don't need to be shown diagrams of conjugation systems, or even to know that there's any such thing – they just soak up the skills required to operate one effortlessly.  If you try to fob children off with a grammar‐deficient pidgin, which doesn't quite satisfy this thirst for a fully functional mother‐tongue, they're liable to improve on what they're given, getting together and fleshing it out into a creole language.  In one recent famous case, the pupils at a Nicaraguan school for the deaf bypassed their teachers' clumsy efforts to get them to lip‐read Spanish and instead developed a more satisfyingly useful “sign creole” all of their own.  But the knack for automatic language acquisition seems to fade away over time, which explains why picking up an accentless fluency in a new language goes from being child's play to almost impossible somewhere around puberty.

More controversially, many linguists suspect that babies have biassed expectations about what sort of grammar they're going to be presented with – innate prejudices that make them more effective at learning a human language than you'd expect a hypothetical grammar‐detective to be that was entirely reliant on empirical evidence.  Infants down the ages have been given the task of deducing how it all works; the ones that picked it up quickly could get on with discovering what “wolves” were and how to wheedle treats from grandparents.  We can expect to be descended from the ones whose intuitions happened to make them fast human‐language‐absorbers, a filter applied consistently over so many generations that the tendency to make good guesses may have become partly instinctive.

(As it happens the linguists who believe most strongly in such an inbuilt mental module tend to present it as if it was a sort of Language Acquisition Wizard –

Looks like you're trying to assimilate a native tongue.  Is it:

[ Cancel ] [ < Back ] [ Next > ] [ Finnish ]

…But it doesn't have to be anything like that, honestly!)

Now, it's clear that whatever hereditary component there might be in this putative language instinct, it produces no variation in the grammars that modern babies from different parts of the globe are capable of handling (look at the ethnic diversity and linguistic homogeneity of the USA).  So if it really happened, the process must have had three phases.  First was the phase of mutant mind powers – the period of competition between types of brain, and types of language, that provided genuinely different levels of communicative functionality.  Then the “winning” system became uniformly present in the ancestral genome.  By the time our sapiens forebears started scattering out of Africa (before 50,000 BC), we must already have been into the third phase, in which the detailed accommodations between neuroanatomy and grammar were always in the quicker and easier direction of our languages evolving to suit our brains rather than the reverse.

Unfortunately, any coupling between human neurology and human languages makes things difficult when it comes to theoretical discussions about logically possible alternative kinds of language… the self‐evident superiority of our way of doing it may be no more than a natural bias in our intuitions, like the one that made vision seem such a simple thing to handle until we tried actually implementing it in working software.  If common sense in a vacuum is untrustworthy, maybe what's needed is experimental constructed languages with designs that don't match our prejudices, so that we can evaluate whether they're plausible alternatives that need to be taken into account or whether that aspect of language was always a necessary “forced move”.

PROTOLANGUAGE – Origin of Speeches

Are all present‐day languages descended from the language of the earliest hominid to start talking?  Well, no, some of them, like the Nicaraguan case I mentioned, have pedigrees stretching back only decades or centuries.  But ignoring oddities like that, which may well have been rare until historical times, yes, “monogenesis” is plausible enough.

There have even been attempts to reconstruct a vocabulary for “Proto‐World”, the hypothetical shared ancestor of all modern human languages.  Now, it's true that when you compare Icelandic, Bengali, and most of the languages in between, you can can trace a family tree linking each of them to a single stem in about 5,000+ BC known as Proto‐Indo‐European; and there are tantalising hints that Proto‐Indo‐European and its immediate neighbours in Africa and Asia are all members of a superfamily.  However, orthodox linguists don't expect anything solid to come out of genealogical speculations this remote; inter‐language relationships more than a few millennia old are hard to make out in the background noise of coincidental resemblances.  Language evolution can be remarkably orderly in the short term, but in the long term it's a process of unpredictable reshuffling and reorganisation on every level.  Once you're looking far enough back to unify the dialects of Wagga Wagga and Ouagadougou, any pattern in the static is guaranteeably a hallucination.

Mind you, whether the optimists working towards “Proto‐World” get anywhere or not, their target isn't the primordial hominid Ursprache.  We can reconstruct Proto‐Germanic from its known daughter‐languages (Old Norse, Gothic, Old English, etc.), then use Proto‐Germanic and its siblings (such as Hittite and Proto‐Iranian) to deduce things about Proto‐Indo‐European, and so on up the wobbly ziggurat of conjectures.  But the furthest back this can possibly take us is the Most Recent Common Ancestor of the world's attested languages (that is, the latest language that all others can trace a line of descent from).  And that's a retrospectively conferred title, not a mark of any special intrinsic property.  Imagine that Proto‐World first divided into a thriving and diversifying eastern branch, which overwhelmed and replaced all rival language families, and a less successful western branch, which today is represented by only one surviving leaf: the secret ritual tongue of some tribe in the Kalahari.  If that language dies out, so does any hope of reconstructing its origins – the furthest back we can get is Easternese, which claims the title of Most Recent Common Ancestor.  And so on; as minority languages disappear, the MRCA moves closer and closer…

Or then again it's possible to imagine scenarios that would push the MRCA back into the distant past.  The current consensus picture of our evolution features waves of increasingly human‐like species, from Homo habilis through to Homo sapiens, radiating out and supplanting the existing pre‐human populations.  Some of these waves may have involved languages as well as peoples being driven to extinction.  Or not – like the Manchus (in China) or the Normans (twice, first in Normandy and then in England), the invaders may have ended up adopting the local vernacular themselves.  In which case, some current human languages might be descended from non‐human languages!

We don't have any real idea how unlikely that scenario is, since it isn't clear even in order‐of‐magnitude terms how long ago genus Homo developed a rudimentary linguistic capacity.  If brain size is anything to go by, the big inflation in processing power started in 1,000,000 BC.  There's evidence not much later of Homo heidelbergensis flint‐knappers sitting together in a circle, as if they were chatting as they worked.  So even 100,000 BC may be a low estimate – by then sapiens babies had been around for a long time, with throats and speech centres pretty much equivalent to modern ones (and presumably similar instinctive urges to use them).  By around 10,000 years ago there are early stirrings of what would soon become writing.  And yet I still occasionally find reference books dating the origin of language to about that late… an idea that leaves me speechless, along with all my sub‐Saharan forefathers.

NEANDERTHALESE – Conlang of the Cave Bear

Oops, now that I've mentioned the N‑word, I suppose I'd better perform the ceremonial exorcism of popular misconceptions.  Forget the cartoons of shambling misshapen apemen dragging their willowy mates about helplessly by their hair.  Neanderthals had sturdy, muscular physiques – and that includes the Neanderthal women; they may even have gone out on mixed‐sex hunting parties.  These so‐called “stone‐age cavemen” made most of their tools out of wood, not stone, and lived mostly in camps, not caves.  They stood fully upright, had brains as big as ours (or bigger, albeit with smaller frontal lobes), used fire, buried their dead, and lived as successful big game hunters in social groups that took care of their sick and injured.

That's the current state of palaeontological knowledge about Homo neanderthalensis; but what follows is a fantasy – a sketch for an incompletely developed language, inspired by a question that's been bothering me.  Given that we decided to build our sentences out of prefabricated lexical tokens that were bursts of articulated noise (“words”), why did we go for the apparently overengineered option of constructing these tokens themselves out of sequences of smaller building‐blocks (known to linguists as “phonemes”, but you should get my point if you only think of them as “letters”)?  In the long term, it's a beautifully extensible scheme, but who was planning for the long term?  At the early stages, it would be simpler to assign semantic functions directly to the different distinguishable monosyllabic utterances.  And from there, well, there's at least one alternative path that might have been followed.

Please note: from this point on, it's speculative fiction.  I'm doing my best not to say anything that there's solid evidence against, but I don't want anybody to believe that it's true!

First, some introductory notes on the Old Stone Age.  Neanderthals were no “noble savages”, and their lives were far from photogenic; they tended for instance to regard any trespasser as a potential meal.  However, many of the unpleasant practices we think of as “primitive” and “barbarous” – torturous initiation rites, blood sacrifice, et cetera – were thought up by Homo sapiens.

Two things Neanderthalese definitely wasn't:

  1. Ungrammatical – People don't say “I ain't done nuffink” because they're throwbacks; they speak like that because they're successfully following the grammatical conventions of a speech community where (as in Ancient Greek) so‐called “double negatives” are mandatory.  It may be a community that's despised by the ruling classes, but modern social stratifications have nothing to do with palaeo­anthropology.
  2. Childlike – Early humans didn't evolve speech via the same stages as a child developing the ability to talk (any more than there was a stage when australo­pithecines crawled on their hands and knees).  Indeed, even Neanderthal infants didn't babble in quite the same way as present‐day children.  Physically, they matured faster, but the mother‐tongue acquisition process was slower – almost more like modern adults trying to pick up a second language.

Two things Neanderthalese kind of was:

  1. Pidginish – When human communities with mutually incomprehensible tongues interact but lack the patience for actual language‐learning, what you often end up with is the kind of degenerate semi‐language called a pidgin.  Usually these either die out or (if a community of native speakers develops) evolve into true languages; Neanderthalese was in some ways like a permanent pidgin.
  2. Complicated – Neanderthal languages were the most elaborate artefacts in the whole late‐palaeolithic Mousterian culture, full of exotic sounds, grammatical oddities, and esoteric figures of speech.  They were trying to do all the work of conveying semantic content with an inadequate basic rulesystem plus a whole collection of special‐case extensions; you might compare them to a cobbled‐together software project badly in need of a redesign.

And two things Neanderthalese definitely was:

  1. Strange – There are a number of general rules that statistically tend to be true of modern languages, whether because we've lost some flexibility or simply converged on good solutions.  Neanderthalese blithely ignored some of these “linguistic universals” (for instance it didn't exactly have verbs).
  2. Primitive – Surviving hunter‐gatherer populations may have more words for bird‐of‐paradise plumage than for sub‐atomic particles, but that doesn't mean their languages are “primitive”.  They'd function perfectly well in any physics classroom if you gave them a chance to come up with equivalents to such English technical terms as “Gedanken­experiment” and “Eigen­vector”.  The earliest “archaic” languages, on the other hand, were fundamentally limited in their capacity to express ideas.

When I refer on this page to “Modernese” I'm just using a handy cover‐term for all the dialects of modern Homo sapiens (and Ancient Sumerian counts for my purposes as modern); but the language I describe below is “Neanderthalese” in the truest sense.  It was spoken forty millennia ago in what's now called the Neander valley (near Düsseldorf) by the man known to science as Neanderthal 1 – the original type specimen of Homo neanderthalensis.  One reason for being so specific here is that what I'm describing is the male form of the language!  The subtle physiological differences between male and female vocal tracts were emphasised in Neanderthal society; an avoidance of rhotic (R‑like) sounds was taken to be characteristic of female speech, and while both sexes had a tendency to lisp, “slushy” sibilants were a feminine trait.

PHONOLOGY – The Voice of the Mousterians

The way systems of linguistic sounds work in Modernese is that each language maintains a smallish inventory of discrete building‐blocks of sound (“phonemes”) distinguished from one another by patterns of features.  For instance the English phonemes /z/ as in ZEAL and /s/ as in SEAL are alike in being alveolar sibilant fricatives, but differ in the single feature of voicing: /z/ is voiced, meaning that it involves vibration of the vocal cords, while /s/ is voiceless, meaning that it involves air passing through with no vibration.  The same distinctive feature separates /v/ as in VEAL from /f/ as in FEEL, /d/ as in DEAL from /t/ as in TEAL, and so on.  Other languages have different inventories of building‐blocks which treat different sets of features as important and distinctive; for instance Spanish has just one sibilant that may be pronounced voiced or voiceless as convenient in different contexts (and the same was true in Old English).

Neanderthalese phonology wasn't like that – instead the distinctive unit to which features were applied was syllables as a whole.  Each syllable was a bundle of abstract phonological features, represented here by marking the values of each of eight features within curly brackets (only visible in CSS‐enabled browsers).  To take a simple example: NGA indicates “Nasal Dorsal Plosive Plain Open Mid Neutral Normal”.  You might transcribe the pronunciation of that combination into (Unicode) IPA as something like [ŋɑ̃], but only as a rough guide – even if Neanderthal throats had been exactly like ours, it still wouldn't really be appropriate to think of it as a string of discrete elements (subdividing it into an “onset” phase and a “release” phase is about as far as you can safely go).  Nonetheless, the notation is designed to resemble an ordinary alphabetic script, with the specific feature‐marking symbols chosen to give an impression of the resulting sound; and assimilation rules are applied to prevent some of the more obvious mis­interpretations – for instance NGA is written that way rather than as NKA in order to avoid the implication of a voiceless element.

The Syllable‐Feature System

Here are some detailed examples:

Nasal Dorsal Affricate Glottal Open High Rounded Long = NGY’ÂUH, “woman”
is roughly [ʔʁ̃ɒ̃ː] (even more approximately: “ngghwawng”), a throat‐clearing noise followed by a relatively open vowel, with the whole syllable tense, voiced, nasal, and back‐rounded on a long‐drawn‐out high note.
Oral Labial Liquid Plain Close Rising Neutral Short = VRÉ', “man”
is [ⱱ̥ɘ́ʔ] or maybe [ʋʴə́p̪̚] (even more approximately: “vruh?”), where the onset is a weakly voiced cross between a V and an R, and the release is a rather flavourless vowel‐sound, pronounced on an interrupted rising note.
Oral Coronal Glide Breathy Open Mid Spread Normal = БAI, “family”
is roughly [s̞ʱæ̥] (even more approximately, stage‐whispered “HSHHA”), a weakly strident sound followed by a front open vowel, all pronounced with a broad smile.

These combinations of features can easily give the impression of calling for unusual dexterity in pronunciation, but the truth is quite the reverse.  In a “simple” English word such as, well, “simple” will do nicely, the vocal apparatus has to switch from voiceless to voiced to voiceless to voiced, in perfect coordination with some delicate adjustments in the positioning of the tongue and lips, and with the switching on and off of nasal airflow.  To the Neanderthals, that would be a real tongue‐twister; they were used to whole syllables that either were or weren't “voiced and/or nasal”, with much more room for sloppy synchronisation.

If you add them up, you'll find there were barely five thousand combinations (and “ug” wasn't one of them).  This may seem a limited vocabulary, especially given that many of the potential syllables were unused, but it was enough.  Over the millennia, there were minor variations in what syllables (and combinations of syllables) were used for what purposes, but the phonological system itself was naturally resistant to the constant piecemeal sound‐shifts that affect our languages.

GRAMMAR – Flint‑Axe Syntax

This guide doesn't need a section on the rules of Neanderthalese phonotactics, because their words weren't made of strung‐together phonemes; it doesn't need tables of morphological paradigms, because their utterances were all built out of uninflected monosyllables; and it goes without saying that there's no need for a section on their writing system.  This fits the stereotype well enough; they weren't as linguistically gifted as us, so it makes sense that their language should strike us as having some holes in it.

But if anybody came here expecting their grammar lessons to amount to a handful of easy rules, then their prejudices have led them astray: this was, in a way, the one thing they had too much of.  A description of the essential rules of basic Neanderthalese would have many more pages than an equivalent summary for English, because instead of having a manageable number of highly generic sentence‐building rules (like “subject, verb, object”), the language was made up of a patchwork of formulas that applied only to subsets of the vocabulary.

People learning English still have to memorise a good few such rules – consider for instance phrasal verbs such as “give off”, “set off”, “show off” (or “give up”, “set up”, “show up”).  The verb and its adverbial‐particle companion may each contribute something to the sense, but they do it in a vague, arbitrary, and unpredictable fashion.  Sometimes patterns are detectable, such as that you can “eat something up”, “chop it up”, or “use it up” (with a shared figurative element of complete consumption)… but if you try to extend that pattern you'll find that you can't “spend something up”, “damage it up”, or “devour it up”.  There's no particular logic to it; those forms simply don't happen to be on the list of generally accepted usages.  These days such irregular subsystems are uncommon enough to be seen as exceptions, but in the Europe of 40K BC they were (so to speak) the rule.  Children learned each grammatical pattern along with the particular set of lexical items needed to give it a context, so the “mental grammar” of the language that they built up might almost equally well be called a “mental phrasebook”.

Underdeveloped structural features

Of course, even today there are practical limits on the complex abstract parse‐trees we can process on the fly, but unlike the Neanderthals we routinely push these limits with everything from convoluted written legalese to jokes (“time flies like an arrow; fruit flies like a banana”).


This category was by far the larger, partly because content‐words could be built up as arbitrarily long polysyllabic compounds whose meanings were loosely determined by the semantics of the component parts.  Take for example the previously mentioned syllable NGA, which had the basic sense “blade”, but was also the normal word for “(finger/toe‑)nail”, “claw”, “horn”, “thorn”, or “sting”.  Given an appropriate frame of reference it could mean “flame”, “fang”, “lightning‐bolt”, “icicle”, “glancing look”, “interrupting noise”, or “cramp”; and then it occurred as an element in compounds such as NGA–J‘Ä meaning “leaf” (literally “blade branch”).  A further factor that helped Neanderthalese to fit its vocabulary within the restrictions of its syllable‐feature system was homophony: there was a second, unconnected word NGA that might be translated “settled weather”.

By the way, please don't interpret all these cases where Neanderthalese failed to draw simple dividing lines as implying that Neanderthals themselves were confused thinkers.  They relied on verbalised reasoning much less than we tend to when making decisions, so they had no difficulty telling lightning‐strikes from nettle‐stings!

Not all members of the content‐word category were things we'd translate as nouns, though.  In particular a great many of them were (also) capable of functioning as verbs, if not in the same sense as words like “destroy” or “undergo” (which are inherently transitive verbs); they always required support from appropriate structure‐words, both to force the actively verbal interpretation and to tie down the roles of any further arguments involved.

The not‐quite‐nounlike content‐words also included NJËI, the nearest thing to a personal pronoun: depending on the context it could mean “this”, “that”, or “I/you/he/she/etc.”.  Most of the time pronouns were not used – after all, general terms like “place” or “woman” were already monosyllables – but a special word MBV‘AU' was available for referring back to a recently mentioned argument; so this could function as a reflexive pronoun, among other uses.  Then there was PFÀ, which marked an accompanying expression as forming a name, or something equally definitive; thus PFÀ NGA–J‘Ä “(the guy known as) Leaf”, or PFÀ KÈ' “the eldest (unique senior)”, etc.

Neanderthals had no use for numerals.  Sure, they could recognise the difference between four horses and five horses, or between four tally‐marks and five tally‐marks, but they didn't routinely see “fiveness” as a nameable property shared by all quintets.  They did have ways of marking plurals: a word could be repeated to give the sense of multiplicity or iteration (NGA–NGA; this is probably the world's commonest pluralisation mechanism to this day).  A special word R‘EU was used to mark complete sets such as a pair of legs or a pack of wolves.


The largest grouping within the structure‐words was the “relationals”, which worked like something between case‐markers and auxiliary verbs.  They could introduce content‐words in the manner of prepositions with nouns, or give them descriptive or verbal senses; when strung together in chains they could interact in subtle and unpredictable ways.

MBÀUH, “use, apply”
Combined especially with words for body‐parts and artefacts.  Some combinations were transparent, like “use blade” meaning “cut”; others require a little explanation, such as “use ear”, which could mean “puzzle out a solution”.
MB’AU', “prepare, work”
Not to be confused with the above; “prepare blade” was “sharpen” and “prepare ear” would have to mean in a culinary sense.
PÂH, “seem” (or “self‐evidently be”)
Could function as a caselike marker introducing things seen, heard, or imagined.  Alternatively, depending on context, it could mean roughly “resemble”, and was used to create (for instance) colour‐terms.  PÂH NGA–J‘Ä (“resemble leaf”) was of course “(be) green”.
PÁU, “approach”
Conveyed motion towards, and was particularly prone to being strung together with other relationals into delicately nuanced combinations.
TSEH, “join, accompany”
Alone, this relational was often used as the preposition “with”; but the combination PÁU TSEH meant “right up to and/or straight into”.
KÀU, “feel, undergo”
Used with – usually after – words for emotional and physiological states, but also weather‐systems et cetera; adding prepositional relationals implied that it was a forecast rather than an objective report.
T‘EI', “belong”
Used to express intrinsic association (“the tiger's claws”, “the tiger's mate”).  It's also translated as indicating possession, except that no distinction was ever made about which way the association went – it was just as correct to say “I belong to a blade” as to say “a blade belongs to me”.

Relationals, while common, were far from the only kind of structure‐word in Neanderthalese.  Many of the others were modifiers functioning rather like English modal verbs (“can”, “should”…) at their least systematic.  For instance, the nearest thing the language had to a tense‐marking system is the contrast between MWÁH, “know, claim (as a fact about the present) that…” and NDZ’ÊIH, “remember, claim (as a fact about the past) that…”.  Other somewhat similar contrasting pairs of particles included VRÁI “stealthily, deceitfully, gradually” vs. NÐAI' “forcefully, obviously, suddenly”, and T‘ÉU' “together” vs. NJAU “apart, separately”.  And there were three different ways of forming a negative, used with different kinds of statements or with different overtones:

VR’ÁH, “deny”
Used with many informational, stative, or descriptive constructions, and/or with the implication that it was controversial.
K‘E, “avoid”
Used with reference to actions, and/or with the implication that it was a good thing it wasn't true.
ND‘EU', “cannot”
Giving an implication of constraint, but also preferred in negative commands.

Another feature of the language that was distinctively Neanderthalish was the “punctuation‐words”, whose equivalents in English may be conjunctions (such as “but” or “so”) or interjections (like the ones conventionally spelled “ahem!” or “mmhmm”).  They worked to tie utterances together, without themselves necessarily having any clear syntactic role; and they often occupied a phonological grey area, being neither regular syllables nor unstructured sound‐effects.  Many of them were normally supported or replaced by paralinguistic cues such as frowning or shrugging.  A typical example is ’A, meaning “well…” or “now then”.  Less typical are the overworked near‐structure‐words NDÈ (“and/certainly”) and NDÉ (“or/possibly”):

an ostentatory utterance: “look, it's a man, and a woman too!”
ostentatory: “look, it's a man or woman”
an emphatic statement: “indeed he was angry!” (literally “and that blood feel”)
a yes/no question: “was it a man?” (literally “or man or deny man”)
a wh‐question: “who was angry?” (lit. “or this blood feel or that blood feel”)
a conditional: “if it's a man then it isn't a woman” (lit. “or man and deny woman”).

EXAMPLES – Towards Babel

Since I'm trimming all this down to something I can explain in a hurry it may seem that I'm contradicting myself, on the one hand claiming that Neanderthalese lacked structure and on the other giving examples of individual constructions which all appear to work perfectly well.  The thing is, those low‐level building‐blocks were carrying all the load.  Neanderthalese sentences were a sort of bottom‐up cooperative effort where you knew what role word A served because it occurred in a formulaic combination with its immediate neighbour, word B, and together they established a semantic infrastructure that made sense of word C; for instance, NGA used alongside a name or kin‐term had to mean “glance”, and be followed by some indication of the target.  If the meaning was unclear it was general‐purpose reasoning from context that was relied on to sort it out – which may seem sensible, but it's not how it works for us moderns!

In our languages, sentences are made of standard tokens that plug into an overarching abstract framework of syntactic structure.  If we ever want to say something that isn't a natural fit for this framework, it's the grammar that takes priority – for instance when you say “it looks like it was snowing a really heavy one last night”, you're inserting dummy‐words (“it”, “one”) not because they refer to particular things in the real world but because there are syntactic slots that can't be left empty.  For a fluent speaker of English, it's a totally automatic, unconscious process in which your native‐language grammatical intuitions bypass everyday practical experience.  If challenged, you're likely to try to rationalise it in terms of it being the weather that's snowing, in much the same way that an Italian might confabulate justifications for trees being male and doors being female.

As a significant knock‐on effect, Modernese tends to be good at accommodating descriptions of unfamiliar or imaginary situations.  If a Neanderthal had attempted to say “my wings have never tried to know the narrow mornings”, the parts wouldn't have assembled into an interpretable whole; but while in concrete terms it may be nonsense, it makes a plainly grammatical English sentence.  We can even go on to argue that since the presupposition that you have wings is false, logically speaking it's a true proposition!

Another side‐effect is that usable sample texts are hard to come by; the widely used ones are full of anachronistic props, alien social phenomena, and speaking parts for dumb animals.  Even that old standby, Genesis 11:1–9, is only partially translatable.  But here's a stab at it:

Neanderthalese: ’A БAI NDZ’ÊIH
Gloss (content‐words underlined): now.then family remember
Rough translation: once upon a time…

They didn't go in for storytelling much, but when they did they followed strict groundrules; this introductory formula is required to establish the context that the following is something I'm not claiming to have witnessed personally.

voice together‐together belong set‐family agreement feel
The full set of families spoke together in harmony

The reduplicated T‘ÉU' here implies a shared habit (or especially a friendship expressed in exchanges of gifts).  R‘EU–БAI isn't so bad as an approximation to the idea of all the peoples of the world; but before I could express the idea that in those days there were no foreign languages I would have to spend some time explaining what a foreign language is.  Otherwise, the obvious interpretation would be that I was saying “in the beginning, male and female voices were indistinguishable”!  Instead I've fallen back on what is I suppose the assumption behind the story: that there were no arguments among them.

set‐family prepare home‐path leave home traverse grow glow‐landscape
The full set of families migrated from the territory of the dawn

Long‐distance travel was usually expressed within the semantic frame of moving through particular families' territorial ranges, but here there are no non‐nomads to move relative to – this isn't a small band of wanderers I'm describing, it's the entire unified world population!  I'm obliged to make do with treating the dawn as a sort of stand‐in territorial term… though oddly enough we're the ones stuck with a primitive geocentric model of the sun “rising” where the Neanderthals instead talked about the dawn in terms of daylight coming to the land.

benefit same seem landscape‐meadow‐gap
They found themselves a plain

Implying that they discovered and claimed it, like, say, a patch of mushrooms, rather than just selected it as a temporary campsite.  I'm not even going to try to deal with the placename “Shinar”.  Middle‐Eastern Neanderthals may have had their own name for the area, but the Rhinelanders certainly didn't.

home same grow
They settled (there)

A construction that doesn't fit the relational‐plus‐content‐word model.

man‐man together he.said
Men said to one another:

What the original says is that some guy said this to his friend.  We might interpret that as meaning that people in general were discussing this sort of thing, but the concreteness is useful – for a start keeping a male speaker saves me having to quote things in a put‐on female accent!

come.on prepare hand‐fire forcefully stone‐stone prepare
“Let's build up a fire to cook stones thoroughly”

Well, the actions are translatable, even if the motivations aren't – mmm, baked stones!  But then they go on to declare their intention to use stones as stones and tar as tar… I give up on that bit.

come.on use stone‐stone benefit family prepare home
“Let's make our home out of stones”

Of course it's impossible to convey “let's build a city”, but this much does work.  One of the reasons I gave this speaking part to the men is that a woman encouraging a man to build a shelter would be understood as offering to share it with him.  Neanderthals had no notion of taboo (or magic) words, as such, but they did recognise innuendo…

come.on prepare stone‐toy approach join fog‐skin
“Let's make a cairn right up to the heavens”

A cairn was something children often built.  NJÁI–NÐ’È could in principle mean “barrier against drizzle” or “sack full of mistaken beliefs”, but what it was used for was “the covering made of clouds”, meaning the sky.

come.on belong brow benefit senior
“Let's make ourselves look superior”

This being a relatively opaque fixed expression, it's hard to tell whether it would be accepted as appropriate or whether it relies too heavily on the existence of outsiders they could impress.

come.on avoid approach landscape arrange apart
“Let's not get scattered across the land”

This can be translated as long as I drop the reference to the whole Earth and its face.  The normal context for this image of “scattering” is “like prey from a predator”, which might suggest that the tower is intended as a sort of artificial tree everybody can climb together; translating “tower” as “cairn” should help clarify that it's intended more as a landmark to gather at.

family remember
(Close quotes)

I've just finished quoting what tradition says people said; this clarifies that we're back in the main narrative.

this fog‐skin family belong leader
There's this guy, the leader of the inhabitants of the heavens…

At this point I'm forced to pad the narrative out a little.  Where used like this, RAUH requires an introduction, and the novel ideas involved call for a bit of background anyway, most easily achieved with an ostentatory phrase.  Then since I'm not actually saying “there he is!” I have to include the pronoun NJËI, in this case spoken without an accompanying gesture.

leave fog leader approach home seem family prepare stone‐toy
The leader travelled from the clouds to spy on the people making the cairn

Time‐tourists please note: from a Mousterian point of view, “coming down for a look” is already hostile behaviour.  The “people” here are strictly speaking “the Sons of Man”, but this is difficult enough without trying to convey a patrilineal worldview at the same time.

leader know
The leader said:

Pronounced approximately “wraw vaahm?” – the following being his report to the folks who'd stayed at home.

voice together‐together belong set‐family agreement feel
“The full set of families speak together in harmony”

A useful recap.

know same prepare incipient this seem
“They are making a start on this (that much is evident)”

The final syllable here serves to bridge from the verifiable facts into the intelligence threat estimates.

or join family hunt feel and family lucky hunt‐hunt feel
“If they start out together to achieve something, then they are sure of achieving it”

I can't get close to “nothing they attempt will be denied to them” (who by?), so I'll cancel out the negatives to get this.  Yes, Neanderthalese equated confidence with competence!

come.on expedition leave fog
“Let's set out from here in the clouds”

Seemingly this particular piece of divine intervention requires multiple spellcasters at close range.

come.on harm/undo voice arrange vine behave vine
“Let's get their speech tangled like undergrowth”

This “X‐MW’EI‐X” construction formed conventional similes with plants or animals.  Whether the analogy of tangled weeds would stretch far enough to convey a concept as esoteric as “confounding their language” is unclear.  My phrasing here is itself a bit tangled, as I have to avoid letting the wrong syllables end up adjacent.  KX‘AU NDZÁH would inevitably be mistaken for KX‘ÄU NDZÁH, a common chained pair of kin‐term relationals meaning approximately “kid brother of”.

now.then cannot family together‐together agreement arrange benefit family
“So as to stop any family parleying with another family”

Again instead of language‐comprehension in the abstract I'm talking about particular uses of communication (though this time I'd better not leave out the women).  There's still the question of what it could mean in concrete terms to cause all this to happen, but, well, I can't help them there.

family remember
(Close quotes)

And we're back to the main narrative again.

apart family‐family benefit leader leave home approach landscape arrange apart
The families scattered from there in accordance with the leader's plan

I'm rather implying that he had his angels lying in ambush.

cannot family‐family use stone‐stone prepare home
The families were prevented from making their home out of stones

The text doesn't state what happened to the cairn once the Lord was undisputed king‐of‐the‐hill.

family know
(And that's why, to this very day…)

The part that relates the story to the narrator's present day has to go right at the end, so I'll drop the closing summary of the plot.

belong landscape unique vine behave vine landscape
The place is called “tangled‐like‐undergrowth‐place”.

Alas, Neanderthals aren't really in a position to appreciate this Ancient Hebrew pun on the name of the city of Babylon.

HOLOCENESE – Scrabble for Survival

What I've just described may be an early language from our perspective, but for Homo neanderthalensis it was a very late language, the product of a quarter of a million years of development, with a syllable‐feature system that had been pushed almost to its limits.  Nimbler tongues would have given them more available values for features such as Place or Height, but Neanderthalese simply didn't change quickly enough to bring a short‐term return on investment in that sort of physiological specialisation.

And in fact it was already too late for them.  The equatorial branch of the genus Homo had already made the organisational leap to a segmental phonology.  This opened up a whole range of new ways for grammatical systems to function, smoothing the way to a gradual escalation in the part speech played in their mental processes; and new ways for languages to mutate, encouraging a rise in cultural diversification.  Eventually, people were singing love songs, persuading rivals to back off, telling jokes about the stupid accents those hill‐dwellers have, concocting tall stories about the good old days, and solving problems by internalised symbolic thought.

So by 40,000 BC a bunch of lanky, baby‐faced, nattering Africans were beginning their conquest of Eurasia, spreading along the coastline (Neanderthal 1 lived well inland, with the Rhine only a tributary of the Channel River).  Physically the newcomers were no threat to the muscular, weatherproof natives, and there was no great difference in brain size; but their minds had a new adaptability, and their dependence on ingenuity gave them the edge in a changing world.  The Neanderthals survived alongside the Cro‐Magnons for over 10,000 years, and assimilated some of their innovations, but geneticists have found no evidence of interbreeding; hybrids may or may not even have been viable.  [Postscript: That was true when I wrote it, but not long afterwards there were some new developments in genetic palaeo­anthropology.  They don't make much difference to my scenario, but they do suggest that we should be talking about Homo sapiens neanderthalensis.]

However, if it's true that the newcomers had an innate advantage in their aptitude for articulate speech and verbal reasoning, then their takeover wasn't entirely a story of flexibility beating specialisation.  When it comes to language skills, the story is the other way round, with the “new, improved” humans depending on biological adaptations to provide shortcuts instead of doing it all the hard way.  Neanderthal brains may already (I'm guessing) have had language centres predisposed to break down input into meaningful elements, maintain a mental lexicon full of names for things, and recombine words in significant arrangements.  But as language began to snowball, selective advantage began pushing our ancestors towards the development of (weak) innate constraints on the kinds of mother‐tongue they were happiest to learn.  Natural selection wouldn't care how much sense these constraints made, as long as they accelerated language acquisition – the Baldwin effect would just take what it was given by the statistical tendencies in early Homo sapiens languages and turn this known‐working implementation into a standard.  The rules of thumb may have been no more specific than “expect nouns and verbs, expect agents and patients (probably in that order), expect words built out of strings of phonemes”… all things that seem to us like common‐sense necessities for a usable language, but they'd still seem that way even if they weren't.

Unfortunately, third‐millennium‐AD humans are so dependent on the cognitive substructure provided by speech that we're generally oblivious to it.  In fact you yourself may find it hard to imagine that it could have had a role in early human evolution.  If you are such a skeptic, and you think you'd be almost the same person even if you hadn't learned a modern‐human‐style language, then please step forward!  Experimental palaeo­linguistics needs you to sign some release forms so your children can be used in proper laboratory‐controlled tests…  If only the ethics committees would allow that sort of thing we'd be able to find out once and for all whether babies are born with a built‐in Modernese Study Guide module in their heads, or whether they figure out how to apply switch‐reference markers to a headless relative clause just by being immensely powerful general‐purpose rule‐induction engines (but ones that nevertheless struggle with elementary calculations of probability).

Meanwhile, on an island east of Java, another, more distantly related species of hominids may have survived a while longer: Homo floresiensis.  Given that they're reconstructed as metre‐tall reclusive hole‐dwellers, naturally they've been nicknamed “hobbits” – but if the local folklore about “Ebu Gogo” is based on a grain of truth, it implies they may have had language as well.  Alas, the evidence is too thin to form a basis for speculation (it's thoroughly debatable whether the remains are even non‐sapiens); and besides, writing a conlang called Hobbitese would be a good way of attracting a lawsuit.