A Sketch of a Neuro-Linguistic Theory

Below is a brief sketch of a neuro-structural theory of language with a few supporting comments. following that is an outline of a program for exploring questions in historical linguistics. If similar work exists or you have thoughts (critical or otherwise) I would greatly appreciate hearing them in the comments.


1. Distinctions occurring universally or re-developing frequently in human language will have physically detectable neural correlates. They should be further investigated by pharmacological disruption of subsystems or review of patients with neurological deficits.


2. The basic unit of human language is the noun. All phrases are noun phrases. It is difficult or impossible to derive coherent information from isolated non-nouns, except for direct sense impressions ("green", "loud"). This has strong implications for cognition.

2-1. Chemical and electromagnetic imaging should eventually reveal nuclei or
networks of cells corresponding to specific nouns that when activated by an
utterance, activate earlier than other words in an utterance, regardless of
language word order. Children learning a language produce nouns first.

2-2. While nouns are often grouped into genders or other categories based on
some attributes, these are invariably irregular, and it is better to
use neutral terms like as noun class. Nonetheless frequently re-occurring
categories will have some neural correlate (for example, animate/inanimate
oppositions).

2-3. Possible problem for the theory: supposedly the Athabaskan language Hupa
of Northern California has a very limited number of nouns (a few hundred)
and the language is somehow constituted primarily by verbs [reference to
come]. This is easily the strangest thing I have ever heard about a
language and is reminiscent of Berkeley and Borges [reference to come].
However, I believe there are no native speakers left, I've not yet seen the
primary source and facts I've been able to gather are scanty. My hunch is
that either the primary source was a century-ago grad student who got a
little excited or that the primary source has been misinterpreted. However,
if this grammar is accurate, my noun-based theory is undone unless there are
somehow basic neurological differences between the Hupa and the rest of
us. Further militating against such radical innovations are the fact that
no other Athabaskan languages demonstrate such alienness, even those also
located on the Pacific coast like Port Orford and Tlaskanai [reference to
come.] This is a critical point of analysis for the theory and if any
speakers do exist, it would still be worth doing neural and genetic
investigations to see how they differ.


3. There are primary modifiers that directly modify nouns, in traditional grammar referred to as verbs and adjectives.

3-1. Primary modifiers encode information relating to the noun they modify
(number, noun class, case, time and intention). In many languages time is
encoded on adjectives, confusing to speakers of Western Indo-European
languages (e.g., Japanese, Mohawk).

3-2. Beyond their morphosyntax, the only distinction within the category of
first order modifiers is semantic, i.e.g, whether the modifier can mediate a
relationship or property between nouns. If so, these are called transitive
first order modifiers. If not, these are intransitive first order
modifiers. In English, adjectives are intranstive primary modifiers that
cannot be encoded with time information.

3-3. Even in languages like English or Spanish where there is a class of primary
modifiers not thought of as encoding time, there is a high degree of
interchangeability to the point where the adjective-verb distinction is
unclear. In Western Indo-European languages these are participles. It is
not possible to distinguish whether the last word of "She is finished" is an
adjective or verb because the categories are unnecessary.

3-4. Reflexivity is a form of ergativity; in languages where both exist like
Greenlandic, ergative blocks the reflexive morpheme. [reference to follow]
If there is a "native state" of languages with respect to ergative-
absolutive or nominative-accusative alignment, it is that all languages are
actually stative-active and contain both systems but where they are marked,
one is much better developed than the other. The alignment system is always
an extension of the animate-inanimate system.

3-5. Problem for theory worth investigating: relationship and storage of
abstract terms in terms of senses. To borrow and abuse terminology from
analytic philosophy, if we understand "cat" as the primary "anayltic"
element and primary modifiers are parasitic on the noun's concrete
qualities, then in neurological terms, "cat" cannot merely be a network of
primary modifiers - or they would be the elements. Note again that for any
materialist theory of language there must be some physical collection of cells in some
state in our brains that produces the semantic experience of "cat"; how it
is arranged and constituted relative to these other terms is the question.


4. Languages have secondary modifiers, in English called adverbs, which can modify primary modifiers. Their marginal importance is highlighted by their having little or no morphology, fewer or no rules about order in the sentence, and being often oddly affixed with cognition- related terms ("-mente" in Spanish, "-wise" in English).


5. There are non-content logical operators that mediate relationships between nouns. They are especially incoherent in isolation and are in fact always particles, even in languages that are not considered agglutinating.

5-1. There are two classes of non-content logic operators, those which can be
classified as those which give information about relationships between nouns
(in English, prepositions and conjunctions), and those which do not (in
English, articles; in Austronesian, focus markers, but they are actually the
same word category). Logic operators which do not provide information about
relationships between nouns often provide information about the importance a
speaker places on a noun.

5-2. Logic operators are often marked to agree with the nouns they modify.

5-3. Belying their close relationship, logic operators from the two classes
frequently merge; this could be better support if this is shown to
statistically occur over time more often than their ancestor words' use
together and their phonology would otherwise encourage.

5-4. Like other non-noun, non-primary-sense-datum words, these words are also
meaningless unless they are attached to a noun. In that sense they are like
particles, although they can have active morphology.

5-5. Languages with rich noun morphosyntax (particularly for grammatical role,
i.e. case) will have a far lower use of non-logic operators. To wit, the
parallel and independent developments of increased use of prepositions and
the institutionalization of articles in Western Indo-European languages
(Old English to modern English; Latin to Romance).

5-6. A physical correlate of operators is found in EEG studies. Normal English-
speaking subjects show a smaller ERP on reading prepositions than on reading
nouns, presumably because greater resources are required to recall the nouns
which contains extensive learned, networked sensory content. This
difference is not evident in schizophrenic English-speakers who must expend
the same electrical effort to recall prepositions as nouns. [Reference to
be inserted.] Notably, schizophrenics are grammatically intact but
semantically deficient, in terms of logical relationships of words, word
choice, focus and direction of discourse.


5-7. Language-deprived individuals retain an ability to produce discourse with
adequate logical word relationships, word choice, focus, and direction;
however they are persistently unable to learn to use logic operators in the
correct orientation to nouns, sometimes using them adjacent to primary
modifiers ("the ran"). [reference to follow] EEG studies are predicted to
show a normal lower-than-nouns recall of operator words (but this study has
yet to be done).

5-8. Schizophrenic speakers of languages with rich focus and topic-marking
systems are excellent cases for this theory and others which seek to
investigate the relationship between language and cognition, because
logical relationships in the speaker's cognition are more exposed. Washoe
has a famously complex topic-marking system [reference to follow] but
likely has no remaining native speakers. Tagalog or other Austronesian
languages may be good alternative candidates.

5-9. It is logical operators that allow recursion. Operators can place phrases
in subordination as a primary or secondary modifier to nouns or primary
modifiers respectively, or as equals (we call these operators
conjunctions). Studies of the spontaneous production of recursion in
schizophrenics may be useful.


6. Non-content non-operator terms (exclamations and hand gestures) can be thought of as products of the autonomic nervous system; they can be trained by their utterance does not carry semantic information though they may be rich in social signals.


7. Personal pronouns in this framework are similar to their understanding in other theories. Their repeated re-development in human languages in similar roles and their marking for number, gender, grammatical role, and social status reveals underlying categories in human cognition that must have (eventually) measurable physical correlates in the nervous system.


8. Word order - word order in any individual is the result of inductive learning; i.e., raise a child speaking only English with VSO order, the child will produce VSO English (hypothetically; experiments testing this hypothesis would be informative).

8-1. There are languages that resist classification as having a specific word
order (not just like English where discursive functions change word order;
normally English is SVO but we utter OSV statements for contract - "I
don't like spaghetti, but linguini I like." - and are best described in
terms of statistics, for example Tsimane. [M. Gurven, personal
communication]) This should be troubling to strongly rule-based word
order generativists.

8-2. The current theory predicts only that there will be a general statistical
tendency across human languages for words to occur temporally earlier in
sentences based on how basic they are to cognition. Consequently
noun-before-verb orders will be more frequent, and nouns will more often
than nouns precede first-order modifiers and particles.

8-3. Investigating the relationships between most permitted word-orders is not
necessarily productive (i.e., SOV languages tend to have postpositions not
because of any structural necessity but because there is a plurality of
SOV languages due to the importance of nouns, and more languages have
noun-initial phrases, so statistically we should expect lots of SOV and
postpositional languages; similarly there is little information in a
language like English that has noun before transitive first-order modifier
but noun after most intransitive second-order modifiers). Much more
interesting is arrangements that are universally forbidden. Why are first-order modifiers never in
any language further from the modified noun than logical-operator phrases
acting as first-order modifiers? [Reference by R. Morneau to follow]



Underlying the whole program is that the point of studying language is two-fold: first, to understand human cognition better; and second, to illuminate events in prehistory, which questions can also serve the primary goal. There are historical questions that are interesting for both of these reasons.


Some Investigations With a Novel Historical Linguistics Genetic Approach

A. An approach to determining genetic relationships between languages using phylogenetic methods on morphosyntax rather than phonology as with the comparative method may be fruitful. Morphosyntax must maintain some level of internal consistency and is constrained by neurology, so it necessarily a more conservative element of language and may enable reconstruction of relationships in deeper time than is possible with phonology-based methods, following the call of Nichols and others to focus on critical languages in the hopes that patterns of prehistoric migrations will be illuminated. Once the genetic patterns of well-attested languages have been reproduced by this method, it can be pushed further back for questions as follow.

B. What, if any, are the relationships between the current language families of Eurasia? How do these compare on the large scale to the genetics of the speakers of those families and the timing of the spread of important memes as shown by archaeology?

C. What, if any, are the relationships between the language families of North America? Given the the likelihood that Native Americans are descended from an isolated population of roughly 20,000 that survived in Beringia and began their diaspora near the beginning of the Holocene,[reference to follow] it is likely at the very least that they spoke similar languages at that time, and that deeper connections to northeast Asia exist and could be elucidated.

D. A new approach to the sprachbund problem - can a morphosyntax-based genetic method distinguish between areal (lateral) effects and genetic descent, thus illuminating contacts in prehistory? It's well-known that morphosyntax is the last feature to be borrowed between languages in contact, possibly again because of the need for systemic consistency. After training such a system on the well-attested Balkan sprachbund, it could be applied to investigate possible Uralic contacts with Germanic as the motivation for Germanic's innovations relative to Indo-European, as well as the sprachbund identified in the south Cascades by Delaney.[Reference to follow.]

E. A quantitative approach to model language development over time in the Oceanic branch of Austronesian. The mostly isolated languages of Polynesia can be to linguistics what the Galapagos were to Finches for Darwin. Estimates of population size over time, travel between islands, and internal constrains of the languages' structures can be built into a model that predicts language innovation rate and type.

F. Grammatical simplification of imperial languages. Latin's grammar simplified quickly during the Empire, and the Chinese languages were prefixing fusional languages as recently as twelve centuries ago.[reference to come] Is this a feature of all languages of multiethnic empires, i.e. did pre-classic Nahuatl require more grammatical decisions-per-syllable than the Nahuatl of Moctezuma's day? What is the mechanism (current theory, non-native speakers moving into the polity introducing imperfections through contact via intermarriage and trade).

G. What do the patterns of language evolution over time reveal about human neurology? What can we say about the biological basis of cognition based on the rate of recurrence of certain categories or grammatical rules? Bickerton [reference to come] began the investigation of this question in the context of creoles. With reference to creoles but more abstractly, are there generaly principles of what structural features survive or are innovated when there is a collision between two systems of rules that have some requirement for internal consistency? (e.g. lateral transfer of genes in biology; rule-based productive memes, like music, for example the blues and rock scale in the West which redeveloped pentatonic scales like those used by most humans)?

H. Experimental investigations of the neurologically-mandated structure of language. There are structures in human languages that, while possible and logically consistent, we never see. Attempting some objective measure of complexity, take two groups of subjects and teach each a constructed language (conlang). One group learns a language of that mimics real grammars, the other a conlang that contains never-observed patterns. Use the same phonology and control for complexity; see whether there is a difference in error-free production.

I. Experimental investigations of the impact of grammar rules apart from phonology. Take monolingual speakers. Again teach them conlangs. One conlang has a grammar identical to their own; another has a grammar imitating a real language of equal complexity. See if there are savings, i.e. if speakers more quickly and accurately learn the language whose grammar mimics their own but even though it shares no word roots at all.

J. An interesting and extremely controversial question is whether there are any genetic differences between individuals causing variation in the neurological hardware that the languages are running on. This might easily be true in a trivial sense between individuals; that is, there are genes affecting language use that differ within populations. More interesting but more disturbing is the issue of whether there are language differences in genetically distant populations owing to genetic-based cognitive differences between populations. That is, are there strange twists in Khoi-San or Pygmy languages relative to Indo-European that can be attributed to genetic, rather than cultural isolation? [Reference to come later.] Until recently many scholars put Piraha in this category but a) evidence increasingly suggests the differences in Piraha language usage are culturally-determined and b) we also would be less likely to find a genetic outgroup in a distant spreading zone than in Africa or at least in the Old World [Reference to come later.] This would also be a place to test the lysosomal-storage heterozygote advantage theory of Harpending and Cochrane [reference to come].