An evolutionist approach to language

One approach for analyzing the structure and function of language is to investigate the links between language and the mechanisms that use, produce, and analyze it, and then redirect the analysis effort towards said mechanisms themselves. If we can indeed reduce our analysis of the vaguely defined concept of 'language' to the analysis of physical mechanisms, we could then investigate said mechanisms from various perspectives, and in particular from the evolutionist perspective according to which "a functional requirement can guide the natural creation of a corresponding mechanism".


The linguistic structures
Let us first look at the interaction between the linguistic production mechanism and the environment, and try to identify how the linguistic structures manifest themselves as part of such an interaction.


Linguistic structures

The diagram above assumes the existence of an identifiable 'linguistic production mechanism', and said mechanism is depicted in direct contact with a 'communication substrate' by means of which it interacts with the environment. From this perspective, the linguistic production mechanism effectively acts over the communication substrate, i.e. its inner structure necessarily "terminates" with an actuator positioned at the junction with the communication substrate. In this context, the linguistic structures can now be seen as the set of shapes that the communication substrate takes when submitted to the action of a linguistic actuator.
  • for example, in the case of spoken language, the communication substrate is the air around the speaker, and the set of shapes that this substrate takes at successive moments in time is represented by the instantaneous sounds (phonemes) that can be heard. The linguistic structures, may they be static (i.e. the set of phonemes used) or dynamic (i.e. the way the phonemes are chained in time), are directly reflected in the shapes that the communication substrate takes, i.e. in the sounds that can be heard in the air.
If we look at the linguistic structures (both static and dynamic) from this perspective, they can now be considered as properties of an object (i.e. of the communication substrate), and they are a direct reflection of a "modeling activity " performed by the linguistic production mechanism. Thus, based on this remark, we can now reduce the analysis of the linguistic structures to the analysis of the linguistic production mechanism.


Language in a sensori-motor loop
Based on the remark that the linguistic production mechanism is effectively acting upon, and reshaping, its environment, we can now identify a sensori-motor loop that involves the use of language and that consists of a transmitter, the environment, and a feedback path from the environment back to the transmitter:


The language sensori-motor loop

The diagram above is very general, and is applicable not only to natural language (in all its forms, e.g. spoken language or sign language), but also to any form of activity that involves an interaction with the environment based on a sequence of activities that exhibits linguistic properties.
  • for example, the diagram above can represent a system in which a crow is performing a complex succession of activities required for accessing food (new caledonian crows can even build tools and make use of meta-tools in order to accomplish a goal): in this case the linguistic production mechanisms will not generate a succession of gestures required to articulate sounds, or type a sentence, etc, but rather the succession of activities (the "words") involved in the goal-oriented activity (the "phrase") of gathering the food. Moreover, a similar linguistic model can constitute the basis for even far more common activities, e.g. the succession of activities that a dog performs when he buries (or recovers) a bone.
An important remark to be made at this point is that the individual activities that are being sequenced into a new activity are (usually) themselves complex combinations of other activities.
  • for example, when a dog performs the activity of hiding a bone, this activity involves (a specific sequencing of) other complex activities, e.g. digging and walking.


    A dog's activity of hiding a bone
  • note: any activity that is executed by an actuator, no matter how complex, ultimately has to consist of a succession of simple actions because of the serial nature of the actuator. Also, because most of the common activities are not performed via one single actuator, several parallel successions of actions are usually required (one for each actuator).

The natural language loop
Let us now look at the special case of natural language, where the environment contains a receiver cognitive agent that analyzes, and responds to, a natural language-based message sent by the transmitter.


Natural language loop

Two closed loops are superimposed in the diagram above: the generic loop described in the previous paragraph where the recipient of the message (which is also the feedback originator) is "hidden" in the environment (path: transmitter cognitive agent -> environment -> transmitter cognitive agent), and the special-case natural language loop that explicitly requires the presence of a natural language analyzer at the receiving end of the transmission chain (path: transmitter cognitive agent -> environment -> receiver cognitive agent -> environment -> transmitter cognitive agent).
  • note: the details of both feedback paths are not relevant at this time, and thus they are only indicated by means of their originating and terminating ends.
The form in which the natural language loop is illustrated above allows us to approach the analysis of the linguistic mechanisms from a evolutionist perspective: we can now try to identify the physical elements on which natural language depends, and then determine what evolutionist pressures are acting upon them and how these pressures can shape their functionality (and thus their structure).


Top-level structure of the natural language path

In order to attempt analyzing the language mechanism from an evolutionist perspective, we first need to see if the evolutionist paradigm can be applied to it. This problem now raises the question of what, if any, evolution capabilities does the mechanism of language posses: should we determine that evolution capabilities exist, it then results that various evolutionary pressures (which will be investigated later) can indeed shape the structure of this mechanism.


Top-level evolutionist decomposition of the natural language path
  • note: the above diagram focuses on the linguistic analysis and synthesis structures, and is not meant to exhaustively include all the information and control pathways in the system. Also, because the nature of the interaction between the linguistic synthesis/analysis mechanisms and the rest of the cognitive structures (the 'cognitive controller') is as yet undetermined, said interaction is illustrated by a generic bi-directional information pathway
The image above represents a top-level decomposition of the natural language path into its key components, based on the ways that each component could have evolved: specifically, the green modules are the oldest ones on the evolution scale, the blue modules represent more recent evolutions that relied on the already-evolved green modules, while the red modules are "program modules" that are learned during a system's lifetime and are independent of biological evolution pressures (i.e. they evolved as a result of cultural rather then biological evolutionist pressures). Also, a very important remark to be reiterated at this point is that the actuator and sensor modules need not strictly represent an acoustic interface, but may also be an interface for e.g. a written or sign language; in other words, the mechanism of language can be uncompromisingly investigated from a communication media-independent perspective.


Conclusions
The evolutionist perspective over language as presented in this note allows us to find a way of decomposing the mechanism of language into categories of modules that (could have) evolved independently because of the different kinds of evolutionist pressures they were submitted to. Based on this decomposition, the modules involved in the mechanism of language can be investigated incrementally based on the position of each module category on the evolution time scale. By determining what evolutionist pressures were applied on each category of modules we might gain new insights into what functionalities (and structures) could have emerged as a result of said pressures. Specifically, the linguistic analysis and synthesis mechanisms (coupled with the linguistic skills) can be analyzed as an evolving apparatus that (must have) started from more simple structure(s) tailored for a simpler functionality (e.g. maybe as simple and ancient as the walking sequencing "controllers"), and subsequently passed through a series of adaptations that ultimately allowed the emergence, and development, of natural language.

No comments:

All content on this website is covered by the copyright policy of the AI Project.