CHAPTER 18

Meaning and Communication

18.1 The Nature of Communication

The basic function of language is to communicate. The ability to communicate, however, is not confined to the use of language. We communicate with others by means of facial expressions, body positions, drawing attention by pointing, and so on.

We all have a mental 'model' of our environment. It includes a representation of the physical objects around us and their likely behaviour. Our immediate sensory experiences are changing continuously but our model, being an interpreted version of our sensory images, is stored in episodic memory and is relatively stable. This model enables us to take appropriate actions and, for example, to sit on a chair which, at the moment of sitting, is behind us and therefore out of view. It enables us to walk about in a completely dark room without falling over the furniture, provided we have had a prior look at the room. It would be a confusing world to live in if our model of it changed with our changing sensory perceptions.

Along with the model of our immediate surroundings we have a more general model of the world at large. The experience on which this model is based is acquired over a longer time span, and the model itself is stored in semantic memory. It is 'general knowledge'.

When we communicate we enrich the internal representation of others with respect to the environment. If we draw their attention to some phenomenon, a representation of that phenomenon will be added to their internal model of the world about them. If we describe to others a place known to us but not to the hearers, a representation of that place will be added, after due processing, to their world model. For them it now 'exists' in their model, perhaps not as vividly as would be the case if they had seen it for themselves but it 'exists' nevertheless. It is in this way that we can all build up a model of the world which includes countries and places of which we have no direct experience.

Our episodic memory will have a record of the experience of being told facts. Our semantic memory will store the knowledge which results from that experience. Eventually we forget about the actual experience of being told, but we retain the general knowledge: if asked why it is that we are so sure a certain country exists, even though we have never visited it, we are perhaps unable to recall the many occasions on which we encountered indirect evidence of its existence.

The things we communicate about are not confined to simple facts which people might have observed for themselves. If we describe something in a story, our audience will construct a model of a fictitious environment which will enable them to understand the significance of the actions in the story. They will enrich the fictitious environment with their general knowledge of the world. They will be aware, however, that the model is fictitious. Humans are able to handle many models simultaneously without confusion.

Communication is also concerned with transmitting information about our own emotional states. If we frown, others will be able to enrich their model of their immediate surroundings with the knowledge that we are unhappy about something. This makes it easier for them to predict our likely behaviour.

18.2 Symbolic Communication and Concepts

We can tell a story by means of a series of isolated words. For example:

DOG. MAN. BIT. HOSPITAL. RABIES. TEST. OK. RELIEVED.

On reading these words, the reader can extract from semantic memory the concepts associated with each word. Each concept will have a large number of associated 'scenarios' or schematic episodes. A 'hospital' for example, is more than a large building. The representation includes the scenario of people being injured or falling ill, being taken there in an ambulance, being treated, and finally, it is hoped, being cured. It includes the representation of doctors and nurses etc.

Each concept labelled by words in the list above is the centre of a similar cluster of scenarios. The concepts associated with each word will differ from person to person, but will contain a core of similar material. For example, the word 'hospital' has for each of us particular associations concerning particular occasions in our past. However, as expla!ned in section 17.3, we do not expect others to share all of these associations, and so when we use the word 'hospital' we do not use these (personal) associations to form the predictions we make about how another person will interpret our words. The common core of associations may be ill-defined, but it is this common core which permits communication. It is a kind of bargain or contract we strike with our fellows. Occasionally there will be misunderstandings when we assume someone has an association which they do not have, but for the most part communication flows freely.

Some of the scenarios associated with a word will also be associated with other words in the list. For example, the concept labelled by the word 'hospital' will be associated with people becoming ill, and the concept labelled by the word 'rabies' will have a similar association. These overlapping scenarios will be matched and therefore 'confmned' as important to the story being told. Others, which have no commonality, will be held in abeyance until the story appears to require their use. In this way the words in the list generate a 'context' for each other which determines the appropriate implications. The concept labelled by the word 'test', for example, will have associations with examinations and schools. Such associations will not have relevance to the story told in the words above, because no other word in the list labels a concept with similar associations. But it does have associations with medical tests and these are shared by the other words in the collection. In this way the reader can construct a series of scenarios which appear to represent the story told by the set of words.




If we deliberately jumble the words, the story is still discernible although with difficulty and with more possibility for ambiguity (e.g. HOSPITAL. DOG. TEST. RELIEVED. MAN. BIT. RABIES. OK). There is more ambiguity because the chronological sequence of events is not represented. The dog might have been bitten by the man, for example, or even by the hospital. Nevertheless if we give the most likely scenarios preference to the least likely, the story is still more or less as before. From this example it would appear that word order is an aid to understanding, but is not a necessary element of communication. The main message is carried by the words which label concepts and by their associations. Our ability to fit these word-meanings together in a sensible way, based on our knowledge ofthe world, is what makes communication possible. If we not only sequence the words in chronological order, but provide additional clues to the way the concepts associated with the words are to be linked, the story becomes even more easily understood.

A  dog  bit  a  man.
The  man  was  taken  to  hospital. 
Rabies  was  suspected.
A  test  was  made.
The  result  showed  that  he  was  OK. 
The  man  was  relieved.



The additional words and the grammatical structure of the sentences allow us to link the scenarios together and slot entities such as dog and man into the correct roles. The verb in each sentence and its form (active or passive) helps us to identify the role positions. For example, the entity preceding the verb is responsible for the action (in active voice), and a phrase such as 'a man' indicates that we are talking about one arbitrary example of the class of men, not the whole of mankind. Grammatical structure is therefore an aid to understanding. Placing sentences in a particular sequence gives an indication of the chronolog­ical sequence of events. These aids, however, are very imperfect, and the main reason why communication is possible in this symbolic form is that humans know what is reasonable.

In past years the development of natural language processing was hindered by the idea that the grammatical structure of sentences and their meaning were two separate issues. Evidence for this was produced in the form of examples of sentences which were meaningful but ungrammatical, and sentences which were grammatical but 'nonsensical'. What seemed to have been overlooked at that time was the notion of partial meaning. If a sentence which is both meaningful and grammatical is modified so that it is no longer grammatical, then it has lost some, but not all of its meaning. If the important concept-labelling words in the sentence are replaced by others, with the same grammatical classification but with different (and anomalous) associated scenarios, then it has lost even more meaning, but not all of its meaning. Even if such a sentence appears to have no meaning (i.e. it is 'nonsensical') it is invariably possible to create a context which supplies the missing overlap of scenarios and gives the sentence some meaning.

These considerations lead us to suggest that the grammatical structure of a sentence is a meaning bearing symbolism, and that syntax is a sub-topic of semantics.



18.3 The Construction Kit Analogy

A statement made in communication can be regarded as a 'construction kit' for an internal representation of some aspect of the world, real or imaginary. Some of the words are labels for the constructional units, other words (and the order of the words) carry information about how the constructional units are to be put together.

There is a very close analogy with a child's modelling kit. The kit provides the child with constructional units and with instructions about how to put them together. Individual instructions perform the same role as some words in a natural language sentence. For example, some of the instructions are entirely devoted to part recognition and establish the names for these, e.g. 'The wing panels will be found in the plastic bag. The leading edge is the more rounded edge. ' Certain instructions are devoted to the method for joining these, e.g. 'Glue the butt end of the wing panel into the groove on the side of the fuselage with the leading edge forwards. ' The sequence of instructions indicates the chronological sequence in which the instructions should be carried out. However, if the instructions are lost, it is often still possible to put the pieces together in roughly the correct pattern, provided that the person doing it uses his or her common sense. In the same way we might still be able to interpret the story correctly if the words (all nouns) are jumbled.

18.4 The Importance of Models of Models in Communication

The idea that our internal models contain representations of other people's models is very important. It is this that enables us to predict the likely behaviour of others. Since mankind is essentially a social animal, required to cohabit and cooperate with the most intelligent (and the most dangerous) animal known (man himself), that ability is crucial to our survival.

Knowledge of physical objects can be gained by direct observation, but knowledge of the internal model of another person can only be gained by that person communicating the information to us by facial expression, by other body symbols, and, most important of all, by language. It is not surprising, therefore, to find that language abounds with allusions, direct and indirect, to thoughts, attitudes and emotions. The obvious examples are expressions such as:

I  think  that  ...
I  wish  ...
He  wants  to  ...



They  are  going  to.
She  knows  that  ".



and so on. But this kind of information is not carried by such obvious candidates alone. We are all familiar with sets of expressions such as:

'I am of independent mind.' Tou are hard to convince.' 'He is a stubborn fool.' To some extent these all describe the same kind of human behaviour. The real difference between them is the way they provide clues about someone's attitude to such behaviour, and how that depends upon the person whose behaviour is described. Indeed it is sometimes hard to find a form of words to describe something which does not carry some connotation about the speaker's value judgements. That is not an accident.

18.5 Imperative Statements and Questions

There are, however, even more subtle ways of making hidden allusions to internal models. Consider the three sentences:

The  door  is  open.
Shut  the  door!
Is  the  door  open?



The first, said with a flat tone of voice, is a straightforward example of a 'construction kit' statement. The speaker gives the hearer a kit to construct a representation of a door standing open. The hearer then take such action as he feels appropriate. The second statement is also an example of a construction kit communication, but in this case the model is more extensive and includes details about the speaker's attitude to the door being open. It can be interpreted thus:

The  door  is  open.
I  want  it  shut.
lf  you  do  not  shut  it  I  will  be  annoyed.



The first statement can be made to carry the same information as the second, by changing the tone of voice to indicate the speaker's irritation. The third statement is also a construction kit communication, but in this case it describes the speaker's model and indicates that in that model the state of the door is unknown. The interpretation is as follows:

There  is  a  door.
In  my  model  its  state  is  unknown.
I  would  like  to  be  able  to  complete  my  model.
If  you  supply  the  information  I  will  be  pleased.



The first statement could be made to mean the same as the third by a change in the voice inflection, (that is by adding additional information about the speaker's internal model).

18.6 Summary

This analysis shows several things:

(1) The construction kit theory of communication holds for imperative statements and questions as well as for simple descriptive statements.

(2) Information about internal models is carried in subtle ways by very many statements.

(3) An important aspect of internal models is the indication of emotional state and predicted emotional state.

(4) Any attempt to develop a natural language processing system for the general case, as distinct from a restricted domain such as our micro-graphics world, will be doomed to failure if it does not attempt to represent the internal models of people.