In Part #1
of this series, I suggested that a grammar heavily based in syntax was not sufficiently scientific as a general theory of how language functions.
Part #2 was an overview of how linguistic error-handling processes can add to the reliability and predictability of communication using human language.
Part #3 introduced the topic of conformity - the observation that language users broadly conform to the evolved norms of a language-using society.
Conformity applies from the phoneme level to the word and phrase levels of semantics and syntax. Language users must not simply use the same sound codes: they must be consistent in what they intend to convey by those codes.
Each user of a language will choose a way of speaking that 'feels right'. In case of doubt over a choice of words, or of pronunciation, the rule of aesthetic conformity will be applied: if it sounds right, then it is right. In the long term, etymology and prescriptive use of language count for nothing. If a majority of language users feel uncomfortable with a particular use of language then they will either stop using it or they will modify it. For example, the inventors of margarine insisted that the word be pronounced with a hard 'g', based on its etymology. A majority of English speakers preferred the soft 'j' sound, and that swiftly became the norm.
What does it mean for language use to 'feel' right? I suggest that the psychological basis of conformity is suggested by the famous experiments of Stanley Milgram1. We seem to be more susceptible to influence by our peers and our superiors than we would care to admit. We may or may not be conscious of this influence, but it would account in part for our readiness to conform to the norms of language use which we encounter in our social environment. I suggest that conformity produces a feeling of comfort, and that this leads the language user to an aesthetic view of language: a feeling that some ways of using words are 'good', and some 'bad'.
I suggest that the experiments of B.F. Skinner2 are also of relevance to conformity. These experiments suggest a mechanism by which conformity to social norms may feel rewarding. Our aesthetic appreciation of language may derive from circuits in the brain which reward us for broad compliance with social norms. Compliance can never be precise: as mentioned in part 3, biological mechanisms operate within tolerance ranges. Conformity within a tolerance range applies not just to the sounds and rhythms of a language, but also to the meanings conveyed by the words.
The debate about what words 'mean' has been going on for at least 2,000 years. The core argument is really philosophical: is reality 'out there' in the world, or 'in here' in the mind/brain. I propose a pragmatic approach: each perspective has its use in an appropriate context. The concept of mental models will be covered more fully in a forthcoming part of this series. For the moment, a very brief overview should suffice.
Taking the word 'tree' as an example, I shall attempt to show what, and how, the word 'tree' means. I suggest that the mind, in the first instance compares a second physical tree with a memory of a tree previously seen. Common elements are stored as a mental model of a tree. For each tree seen, the model is adjusted and refined, becoming a representation which includes the essentials of 'treeness' and disregards all irrelevant components of specific trees. The model is a gross abstraction which includes aspects of the appearance, utility and origin of a tree.
This mental model, I suggest, becomes the internal referent of the word 'tree', rather than any specific or prototypical external tree. Each language user assumes that their referent for 'tree' is the same for every other user. When engaging in the use of language, conversers come to know, consciously or otherwise, if their mental models do not conform. Any necessary adjustments to the model are made so that the members of a language-using community exhibit a conformity in their understanding of what 'tree' means.
This mechanism of semantic conformity is sufficiently accurate for a language community which shares an environment which appears similar to all members. As a language community spreads out into new environments, new external referents are found for words. This can cause a semantic drift: it is no longer axiomatic that any two members of the language-using community will have closely matching mental models of an object labeled by a common word.
In a global shared-language community, it is possible, probable even, that mental models, the referents of shared words, will conform less than in a close-knit and localised community. Depending on locality and experience, 'tree' may mean a plant big enough to have a road go through its trunk. It may mean any plant suitable for building or furniture making. It may be a source of fruit, of spice, of rubber or of medicine. In the more wind-swept parts of Patagonia, 'tree' may mean a plant which grows horizontally on rock faces.
The example of 'tree' is given as an extreme case to show the limits of semantic conformity. I suggest that the linking of words as labels to mental models is sufficient to account for differences of opinion as to what particular words 'mean'. A language can continue to function as a single, socially cohesive language for as long as the meanings and pronunciations of its words do not exceed the tolerance range permitted by conformity.
As with phonemes and meanings, so with words combined into phrases and sentences: the combination must conform to the expectations of other users of the language. Through conscious and unconscious cooperation, language users come to share preferred ways of stringing words together. This contextual conformity, as with phonemic and semantic conformity, has a wide tolerance range.
Contextual conformity, in contrast to syntax, requires that words used together have at least one semantic link. The degree to which mental models can link according to appearance, utility and other such aspects determines the way in which their labels can be used. The plausibility of a string of words is determined by the mental models recalled by those words. The following list shows some examples listed in order of reducing plausibility.
He hit the nail with a hammer.
He hit the nail with a brick.
He hit the nail with a shoe.
He hit the nail with a sponge.
In the examples, the plausibility of each statement is affected by the applicability of the utility properties of the specific 'tool'. It is unlikely that anybody's mental model of 'sponge' would include the notion that it is used for hitting nails.
In Part #5, the idea of mental models is expanded to show how such models may give rise to a semantic grammar: a set of rules for linking ideas together.
 Stanley Milgram - Behavioural Study Of Obedience
 B.F. Skinner Foundation - Science and Human Behavior.pdf
Random Reward Schedules and the Ambiguity of Language.
I am grateful to Aitch of the Puppy Linux forums for his many contributions to my thought processes and for proof-reading these articles.