Language, brain and computation: from semiotic asymmetry to recursive rules

A garden of forking paths: searching for structure correlations

We share the view of V. L. Deglin on the problem of closer cooperation of the iconic right-hemisphere reception as empirical reality, and the imaginary fictional worlds with left-hemispheric communication and epistemological functions. We can say that the right hemisphere "supplies" semiotic material to create statements, texts, narrative space [7]. At the same time, despite the popularity of dichotomous descriptions of brain asymmetry principles, modern works recognize the principles of distributed processing of all types of information. In other words, both hemispheres work to append each other. The above means that a significant part of the work of consciousness is reduced to the construction of symbolic models of reality. In the analytical philosophical tradition, the question stands: is it possible to understand the work principles of consciousness through the analysis of a language as a universal sign system?

The semiotic asymmetry principles are supported by the ideas of Y.M. Lotman about the types of texts and communication. The author distinguishes two types of communicative situations:

1. Situations where the objective is to transfer constant information; in this case, the syntax structure of a language is the guarantor of the stability of the information channel protecting the text (or a language expression) against interpretive distortions.

2. Situations where the purpose of the communication is the generation of new information; in this case, the language system is configured to shift the standard values to "provoke" the recipient to vary the interpretation of the message [23].

Y.M. Lotman also indicates the structural and functional similarity of the three types of intellectual objects: the individual consciousness, the text as a dialogic generator of new meanings and human culture as a collective intelligence. Our approach finds the idea of semantic (text) the nature of consciousness most valuable. For Y.M. Lotman, the hemispheric asymmetry has the same semantic synthesis nature of the discrete and non-discrete as culture polyglottism. That is, texts produced by intellectual instruments (brain, culture) "consist of two (or more) integrated structures modeling principally different external reality" [23].

The integrative nature of cognitive processes and the modular structure of the gnostic sectors of the brain has been confirmed by modern neurophysiological research [20]. Cerebrum functions due to simultaneous interaction of projection, associative, multimodal and motor areas. In clinical laboratory tests have confirmed that the impact on the left hemisphere of a patient artificially complicates the system relationship of the world of things, despite the fact that the impact on the right hemisphere simplifies social relationships and emotional context of communication.

Modern neurocognitive research provides for an interesting hypothesis of the language being a result of brain relational network (this idea traces back to F. de Saussure and L. Hjelmslev). This hypothesis is supported by factual material of two types: linguistic and neurophysiological. Network relational models graphically present ways to overcome the semantic ambiguity of a natural language. For example, S. Lamb refers to the network structure of multiple interpretations, which are implemented by connotative relations, context and conceptual prototypicality. For example, the phrase "Put it on my bill" contextually is quite appropriate in a bar, but if it is pronounced in the same bar by a duck character, an alternative arising due to polysemy ("Put it in my beak") creates a pun and a comic effect [22].

In cognitive linguistics and computer semantics, the tradition of "breeching sentences" analysis (garden-path sentences) deals with the properties of the conceptual relationships of elements. S. Pinker, referring to J. Fodor, provides a number of examples, the meaning of which a automatically deduces by one single route of a "tree". While the computer, apart from the only correct interpretation, builds four more absurd but absolutely correct grammatical ones.

For example:

Time flies like an arrow.

1. "Time moves in a way an arrow would". (The result of spontaneous recognition by a person.)

2. "Measure the speed offlies like you would measure that of an arrow ".

3. "Measure the speed of flies like an arrow would".

4. "Measure the speed of flies that are like arrows ".

5. "Flies of a particular kind, time-flies, are fond of an arrow " [26. P. 199--200].

Neurophysiological measurements show that the analysis of a similar phrase does not delay interpretation and understanding. That is, the test does not allow even one alternative reading, as it might be in a sentence like this:

"James while John had had had had had had had had had had had a better effect on the teacher".

Without intonation or punctuation marks, this phrase is incomprehensible. With the punctuation, the syntax is automatically organized by identifying verbal groups and objects:

James, while John had had "had", had had "had had"; "Had had" had had a better effect on the teacher. (Where John used "had", James used the "had had"; the teacher preferred "had had".)

Today there are even psycholinguistic schools exploring brainwork during sentence analysis. It turns out that sentences with confusable syntax are automatically restored with the natural analyzer (brain) under the contextual conditions. However, the correct understanding is only possible with the parallel parsing (linear comparison of syntactic and morphological structures), that is, with the possibility of unclear conclusion [12. P. 239--250].

For the proponents of the relational network hypothesis the above examples mean that if there is a "route" for interpretations, there is a system in the brain for conceptual and semantic relationships between the elements of a language system. The neurophysiological evidence of the existence of a language network structure includes the experimental evidence of network components in the columns of the brain cerebrum. S. Lamb identifies the following characteristics in the two types of connections (excitatory and inhibitory):

¦ connections may have variable force;

¦ connections have varying degrees of activation;

¦ connections are enhanced with their successful use (learning process);

¦ nodes have a different threshold value;

¦ thresholds may change over time (as part of the learning process) [22. P. 193].

In other words, the network hypothesis is aimed at finding similar structures in language systems and the material brain substrate. The similarities of hierarchical network structures of neuroanatomical nodes responsible for pattern recognition and speech perception with recursive breeching grammar structures are of utmost importance. A distinctive feature of a neural network is that the higher layers integrate the properties of the lower layers: it is the foundation of syntactic and morphological analysis. Then one can make a bold assumption that conceptual semantics of a natural language also has structural components embodied not only in the abstract categorical forms but also in a specifically organized brain matter. To confirm the neural network concepts, quantitative neuro-anatomical data are provided [21].

Mirror neurons and primacy problem

According to A.E. Kibrik, understanding mind processes based on activity motivation of a language structure is a most promising approach, since the reverse movement ("from thought to language") is complicated by the lack of objective access to the mental structure [19]. However, assuming the hypothesis of uniformity of the brain, thinking and language structure, the neurophysiology is also able to achieve results that are relevant to language studies. The latest works on the correlation of language and brain functions have paid special attention to the phenomenon of mirror neurons and the role of mimetic mechanisms in the formation of language structures. In the context of the problems, the generativists and connectivists, holistics and locationists, gradualists and saltacionists confront each other yet again. The majority of representatives of the polemic parties recognizes a fact of fundamental importance: the meaning (semantics) is functional nucleus of both an abstract language structure and material so brain processes.

Specific attention should be given to the connection of the mirror neurons work with the conceptual and semantic aspect of language activity. J. Rizzolatti and M. Arbib, after the discovery of the phenomenon of mirror neurons by Italian researchers, have developed a large-scale research program explaining the relationship of imitative activity of the brain and the development of a semiotic language system [1; 2. P. 188--194]. They proved mirror neurons existed in the Broca's area and directly participated in the formation of speech skills. Another important finding was the experimental proof of dialogical language consciousness. For the semantic content of a language statement, a skill is required to imitate a number of mental states (hopes, desires, intentions, imagination, emotion, etc.). The ability to reproduce non-verbal components of verbal behavior leads to the socialization of an individual and to the formation of a conceptual picture of the world. The connection of language disorders with the blockade of mirror functions of neurons was proven in the studies of the language picture of the world and verbal behavior of people with autism [27]. It should be emphasized that the formation of logical-linguistic categories, such as subject (Agent), object (Patient), predicate, instrumentative, etc. occurs based on the material of very specific areas of the brain substrate. Recognition and categorization of visual objects is due to the brain's simulation of sensor-motor operations. In other words, some primitive concepts of "semantic atoms" or proto-signs have neurophysiological causality. Moreover, the connection between sensorimotor functions of the vocal apparatus and gestures has been proven experimentally, which confirms the relevance of gestural theory of the language origin [28. P. 140--149]. The key point is that research findings prove that neuron structure cause the principles of mental representations and computational models of grammar.

Therefore, based on processing primary (proto-semiotic) information, the cognitive- semantic approach allows talking about the legitimacy of both directions: from language to forms of thought and from thought to language. In our opinion, such parallelism does not help avoid the question of computability of the semantics of mental representations. Further, we will detail the ontological aspects of semantic computability. If the basis for the conceptual content of statements is formed in neuron links, despite the complexity, semantics should have structural or functional properties of matter: figuratively, both in the brain and in the language. Here the problem of primacy arises again: did the syntax form the principles of the organization of semantic material or vice versa, did the proto-mind as a property of the cognitive nature form the syntactic repertoire of the language plan of expression? The problem of recursive language structure and the methods of information processing by the brain also lies within this question.

Recursive functions and substantial reduction

In the recent decades, the problem of recursion has become one of the central topics of discussion about the structural universals of the language, brain, thinking and even civilization [6]. We may assume that such ideas came after the application of recursive algorithms for programming and detection of such procedures in the bihemi-spheric neuron relationships and language grammar. What is recursion? In general, recursion can be defined as a structural self-similarity or nesting of ideas. In the theory of knowledge, in mathematics, programming, music, geometry, the idea of recursion is to overcome the ontological or metric linearity. In his widely known work, D. Hof- stadter, through the recursive metaphor in music, art and mathematics, explains the nature of consciousness in terms of artificial intelligence [13]. Recursion can be represented textually ("I think, therefore I exist."), graphically (from the ancient pictures of animals eating themselves up to its present form of fractal shapes), logically (the liar paradox), mathematically (Fibonacci numbers or the formula 0 = 1 n! = n * (n - 1)! [where n > 0]). The essence of recursion is reduced to an abstract procedure calling itself or to a component containing a component of the same class.

The problem of self-similar formal structures in linguistic studies has been known since the works of Panini, W. von Humboldt and the early works of N. Chomsky, in which the latter had taken grammar to hierarchical recursive structures. In recent years, the interest for the subject has grown with the publication of the works by M.D. Hauser, N. Chomsky, and W.T. Fitch [11], in which the authors have come to the following conclusions:

a) Recursion is the basis of innate language abilities. (However, along with the sensory-motor system and the conceptual system, the recursion forms a "faculty of language in narrow sense" (FLN).

b) Recursion is the only property of the language inherent only to the human mind. (In other words, all other language properties may be found in animal communication systems)

c) Recursion is a unique feature of language abilities.

d) Recursion is universal, it is present in all languages.

e) Recursion is a unique feature of consciousness.

These conclusions (particularly the contradiction between the first and second maxims) caused heated debate in the scientific world [14]. For example, S. Pinker and R. Jackendorff oppose (b) and argue that human language has a number of specific features that distinguish it from other forms of communication. Actually, the problem of recursion yet again brings us to the problem of Rubicon in the theories of the origin of language [25; 3. P. 300--326]. Also, the authors argue that recursion caters to the "mathematical unit" of consciousness and the so-called "social intelligence", which is contrary to (c) [16. P. 211--225]. The sensational results of D. Everett's research of some languages in the Brazilian Amazon (language Piraha, Muran group) come into direct conflict with (d) [9. P. 621--646]. Finally, T. Gentner and his coauthors have proven that European starlings are able to distinguish successive and inconsistent characters as recursive and non-recursive grammar, which casts doubt on e). Actually, M.D. Hauser, N. Chomsky, and W.T. Fitch also recognize that the navigation ability of animals can be regarded as a recursive system [10].

Perspectives of the minimalistic program

The criticism of HCF's (Hauser, Chomsky, Fitch) article leads the researchers to clarification of the meaning of the term "recursion". Recursive functions in linguistics, mathematics and computer science can be interpreted in different ways. Hulst, referring to Tomalin's work dedicated to the recursion in Chomsky's generative program, demonstrates, how the term has changed its meaning as the theory developed [14. P. 19]. As a result, the HCF's text is based on the recursion interpretation as per Chomsky's Minimalistic Program: in the Minimalistic Program, CHL operations recursively design a syntactic object so that any syntax object (language expression) may be defined in terms of a smaller syntax object. Recursive structures may be found at all levels of a language system, especially in syntax and morphology, phonology, lexical structure, and at the level of logical-semantic connections. Language expression (word, sentence) may be represented as a hierarchical structured object, to which we may apply the rules of inference and the rules changing the inference rules. This explains how the final grammar generates an infinite number of expressions. Recursive mechanisms are responsible for representing language in the form of an infinite number of correctly organized expressions [14. P. 22]. The recursive organization of expressions is implemented under two schemes: "nesting recursion" and "tail recursion" (left- and right-sided).

Example of a "nested recursion":

(1) I met the man [who saw the girl [who left this morning] this afternoon] two minutes ago.

When translated into the Russian language the proposal takes the form of rightsided tail recursion:

(2) [Two minutes ago, I met a man [who had seen that afternoon the girl [who had left this morning]]].

What are other peculiarities of the Minimalistic Program, and how can we relate the formal languages structures it describes to the material principles of brainwork?

In 1993, N. Chomsky wrote another "final" version of the generative grammar theory "Minimalistic Program of Linguistic Theory" with the new updates, changes and clarifications. The American linguist gives increasing importance to the lexical database of a particular language. The core theory remains the same: a person has the innate language ability. Perhaps due to the rapid development of computer technology and the popularity of scientific metaphors "brain - computer", Chomsky offers a minimalistic model based on computation and interface system.

In the Minimalistic Program, a language apparatus is represented by two major subsystems: the computer system and the lexicon. The computing subsystem generates language expression and sends commands to implementation systems. In terms of minimalism, the role of the lexicon is broader than in previous versions of generative grammar. The explanatory power of the minimalist model has shifted mainly to the lexicon, which contains not only the words, their phonological structure, thematic relationships between words (valence), but all formal and signs affecting syntactical processes and different in nature in different languages [15. P. 24].

Chomsky distinguishes two implementation systems: articulatory-perceptual and conceptual-intentional, corresponding to the two interfaces: phonetic and logical forms. A computing module also enables control grammar and relevance (contextual relevance) of statements. We see that in his later version, Chomsky does not separate types of interpretations into individual modules but attributes all types of relationships to the thesaurus and voice implementation. In this version, there is an indirect metaphor that represents language activity as a result of the computation by the biological computer - a brain improved in the course of evolution. The language ontology has been reduced to a set of interfaces and recursive structures.

Conclusions on methodological contradictions