November 25, 2017

Can science be reformed?

An early outline of Goldsmith’s general theory of behaviour or “unified science”, elements of which would be revised in later versions.

From Towards a Unified Science, published in The Ecologist Vol. 1 No. 5, November 1970.

Scientists are coming under increasing attack. To many they are doing more harm than good. Why is this so? Can science be reformed in such a way that it can contribute to the long-term benefit of mankind?

The author argues that this could be so if the different disciplines into which it is at present divided were integrated into a single unified science, but such a task requires a new methodology and a new theory of knowledge.

Can man survive science? As it is at present organised, the answer is most certainly no. Many criticisms can be levelled at scientists but undoubtedly the most serious is that they are so specialised they do not know the effects of what they are doing on anything outside their own tiny field of study.

For instance, nuclear physicists, geared to harnessing the energy of the atom do not know the effects of the radiation emitted during this process on biological organisms. Economists geared to the task of increasing ‘wealth’ are quite oblivious of its effects on our physical and social environment. And so it is with all of them. Their specialised knowledge of a minute aspect of our biosphere and their total ignorance of any of its myriad other aspects, condemns them to waste their talents in devising what are at best gimmicks of little long term value and at worst Frankenstein-like monsters capable of wreaking varying degrees of havoc on our already ravaged biosphere.

It must not be forgotten that when there is a finite possibility that something will happen it is only a question of time before it does. Monsters such as hydrogen bombs, and viruses against which we have no natural controls, and that are capable of destroying most life on this planet, are already at the disposal of would-be aggressors.

In the meantime, scientists are methodically providing governments and industrialists with knowledge to permit the further short-term expansion of our species, increasing thereby the global pollution of the air, seas, rivers and soil by the countless byproducts of agriculture and industry.

In this way they are rendering possible those processes that will eventually transform our biosphere into a biological desert capable of supporting only the most rudimentary forms of life.

What is science all about?

To understand why scientists are doing so much harm and to suggest how this could be remedied, we must first of all determine what science is all about.

Most people would say that science consists in building up knowledge. But what is knowledge? Philosophers consider it is their job to answer this question and an important branch of Philosophy, Epistemology or the Theory of Knowledge, is devoted to this task. Unfortunately its conclusions are far from convincing, mainly because it has such little contact with the disciplines whose generalities it seeks to establish.

Most modern epistemologists are empiricists of one sort or another – i.e. they consider that knowledge is built up by ‘observation’ as opposed to thinking.

Professor Ayer is one of the best known empiricists, and in his latest book he defines knowledge as information that is ‘true’, that we know to be ‘true’ and that we have reason to know to be ‘true’.

As I shall show during the course of this article, this corresponds to no concept that can be of any use in a scientific context, what it does correspond to however, is the way the word ‘knowledge’ is currently used in the English language.

Back to top

Oxford linguistic philosophy

This is consistent with the methods of so-called Oxford Linguistic Philosophy at present in vogue in our Universities, which naively assumes that current English usage (as opposed to Chinese or Basque) can provide information on the functioning of our biosphere instead of on the culture or ‘personality’ of the people who devised it, as Benjamin Lee Whorf showed in the studies of the languages and cultures of the Hopi, Shawnee and Navaho Indians. A language is more than a means of communication. it embodies a complete world-view or ‘model’ of the world peculiar to the society that evolved it, providing an ideal medium for its formulation.

Back to top

How do scientists choose their terms

Such a model is very different from that which scientists attempt to build. For the latter’s purposes, a very different language is required, one that will provide the model with appropriate functional units.

Indeed, for a scientist to build a model making use of classifications designed for totally different purposes, is very much like an engineer building a motorcar out of odd bits and pieces lying about his back-yard which were designed for a totally different machine, such as a typewriter or lawn mower.

The linguistic units of a scientific model must be designed specifically for the function they are to fulfil within a specific model – as is the case with Physics, the most advanced scientific discipline. The atom for instance or the quantum are not terms inherited from our cultural past, nor for that matter do they correspond to observable things as they should do if physicists were to be strict empiricists. They were invented, or more correctly, postulated, to fulfil specific roles in an emergent model of ‘matter’.

Thus there is no reason why we should use the term ‘knowledge’ as it is used today – nor in fact why we should use it at all.

Back to top

What is information?

Instead, let us regard scientists as building up information, a more general term than knowledge. We can talk of the information contained within a gene pool, the nucleus of a cell, a brain or a computer.

These are not vague analogies but specialised instances of the same principle.

Mainly as a result of the work of Shannon and Weaver, communication engineers now use the term in a very precise and measurable way. The information value of a message corresponds to its improbability vis-à-vis the receiver.

We can talk of the information contained within a gene pool, the nucleus of a cell, a brain or a computer. These are not vague analogies but specialised instances of the same principle.

If the message is limited to a certain number of words, clearly all platitudes must be eliminated, and only those things which the receiver is least likely to know, i.e. the most improbable ones, must be mentioned. This notion of information is fine for communications purposes, but it won’t do for our purposes.

To explain what is wrong with it one must look more closely into the behaviour of scientists and other forms of life – in fact at behaviour in general. It is essential to understand in this respect that scientists are not unique. They are not doing anything that other people and even other forms of life are not doing. All require information. Without it they could not adapt to their respective environments.

The difference is that scientists are building up more information than they strictly require to ensure man’s adaptation – information that some people think enables man to control his environment.

To understand the nature of the information that scientists are building up, one must not regard it on its own but as a specific instance of a more general principle.

Back to top

Information in a behavioural context

If we do this, we find that other natural systems do not just accumulate information as empiricists assume they do, they organize it. Data must not be confounded with information. The two are very different. Data is detected, ‘transducted’ or translated into the informational medium or language of the brain and then organized into information.

This process is what we normally call ‘perception’, though empiricists seem to ignore the last two steps and insist on regarding it as a simple mechanical process rather like taking a photograph, instead of a complicated organizational one. It is essential to realise that it is the brain, not the eye, that is transforming data into information and that the brain is really doing the perceiving. Perception is the interpretation of data, and it is for this reason that people tend to see very different things and that perception is in fact so subjective.

Back to top

The measurement of information

If information is a type of organization, then it can be measured in terms of its departure from ‘entropy’, a concept originally developed in thermo-dynamics, which corresponds to a state of total disorder or disorganization such as the primeval dust, where each particle is separate and disassociated. As ‘order’ or ‘negative entropy’ builds up, the particles will organise themselves into ever more complex associations.

Since, in accordance with the second law of thermodynamics, entropy is the most probable situation, improbability must increase as we move away from it.

It is in this way that information is measured in communications theory. Unfortunately this concept of information is not quite sufficient for our purposes. It does not take into account, among other things, the reason why particles join together or why information is built up.

Back to top


The most important and least recognized feature of the behaviour of natural systems, including that of scientists, is that it is goal-directed, or ‘purposive’. This thesis is often referred to as ‘teleology’. It is considered heretical by modern philosophers, mainly because it appears to imply the activity of a little spirit, the so-called ‘ghost in the machine’, and also because of its ‘deterministic’ implications. Today these objections are simply not relevant. How a system is controlled so as to keep it moving in the right direction is explained by the fast growing discipline of Cybernetics without recourse to little spirits, while the fact that all scientific predictions can do no more than state probable developments on the basis of an imperfect model of the system involved, must dispel all possible notions of ‘determinism’.

Back to top


What is the goal towards which behaviour tends? The answer is ‘homeostasis’ or stability, which is the same as saying that behaviour occurs to avoid change, or more precisely to reduce it to the minimum necessary to ensure adaptation to the environment.

This can be done in two ways; either by modifying the environment in such a way that changes are reduced to a minimum, i.e. by building a house, introducing central heating etc.; or alternatively by increasing one’s capacity for dealing with change, i.e. by building up more and more information on which to base predictions that will permit ever more accurate responses.

Back to top


Prediction must be the sole object of building up information, and the capacity of biological organisms to achieve it is quite outstanding.

Thus the development of an embryo in the womb is the result of a long series of carefully programmed moves, each one of which must take place in just the right environmental conditions. This means that the correct sequence of environmental conditions can be predicted.

If I can pick up a cup of tea, it is because at each stage in the seemingly simple, but in fact highly complex sequence of steps, I am unconsciously making, predictions which are being monitored in such a way that each little error in the path my hand is taking is duly corrected.

More spectacular is the predictive ability of baobabs and certain cactuses. These possess countless little pores, which they fill with water when a drought is expected. Since the more water they store the slower is their metabolism, they must store as little as will see them through the drought, which is what they succeed in doing, so much so that it is possible to predict the duration of any drought from the amount of water in their pores.

Scientists are attempting to do just this, except that rather than achieve optimum accuracy they are after the maximum; otherwise they make their predictions by using exactly the same methods.

Back to top

Behaviour based on a model

The first person to describe this method was Kenneth Craik in 1952. He suggested that organisms possessed in their brains a little model of their environment. Every action would be regarded as based on the interpretation of the situation to which a response was required in the light of their particular model. This meant that there was no such thing as trial-and-error learning. If a rat found its way through a maze, it was not as a result of a series of random moves, but of a careful succession of moves each one of which appeared at the time to be the most likely to lead to success on the basis of the rat’s model, as modified by the experience of each successive move.

The scientist learns in just the same way by continually monitoring his model.

His predictions like the rat’s will never be ‘true’, in the sense of ‘absolutely certain’ for two obvious reasons: the number of factors influencing the ever-changing situations are infinite, while he can only take into account a finite number of them, and there must be a time-lag between the detection of the relevant data and their interpretation, during which time the situation may well have changed.

Back to top

Structure of the model

If a model is to represent our biosphere or any part of it, it must clearly reflect its structure.

The biosphere is in fact a single behavioural process, made up of a vast hierarchical organization of sub-processes and sub-sub-processes. All have a number of basic features in common. Thus they proceed from the general to the particular, in accordance with a series of steps that must occur in the right order, and these steps are cumulative and serve to differentiate functions previously fulfilled in a more general manner. Clearly the information organized to form a model must reflect this hierarchical structure.

Back to top

The importance of generalities

In any hierarchical process, the generalities are those that apply to all of its parts. Take an army. The general issues the most general instructions which are then differentiated and further differentiated at each echelon. One cannot explain the behaviour of the whole army in terms of these instructions as at each echelon further information is added in order to adapt them to local exigencies – nevertheless they are by far the most important instructions issued, and to try and understand the behaviour of an army without reference to them would be a pretty hopeless task.

In fact one can say that the more important the instructions the more general they must be, and if one had to build a model using a limited number of variables, their degree of generality must clearly be a major consideration.

Back to top


The object of a model is not to provide a faithful reproduction of a situation, but rather the representation of it which is necessary for specific behavioural purposes, i.e. it must be more like a map than a photograph. Behaviour is required to prevent change, from which it must follow that the model must represent those aspects of the situation which are most threatened with change. If the system is adaptive (and it will not have survived if it isn’t) then it will have developed the means of counteracting expected changes.

In this way the dials on the dashboard of a motor car provide a model of those basic aspects of the behaviour of a car that are likely to suffer change – change that can be counteracted by the driver with the assistance of a garage hand.

Similarly a doctor’s model of the human body will use such variables as blood pressure and body temperature, because a change in their value will affect the body as a whole, i.e. they are important, because such a change is quite likely to occur, and because the doctor knows what action to take to counteract it, at least in theory. On the other hand, a model using such variables as the size of the lobe of the ear and the length of the big-toe would clearly not be of much use to him.

The maximum values for each of these variables within which the system can function properly, can be referred to as its parameters.

It is clear that if we wish to build a finite model of a system, it must represent its most important parameters and those most likely to be affected by environmental change.

Back to top

Information value of the model

The information value of a model reflects its capacity to give rise to the most accurate prediction of the most serious deviations from the systems most important parameters. In other words it must correspond to its ability to interpret messages with a high information value. Such a model must display the following qualities:

  • an optimum degree of order
  • an optimum complexity
  • an optimum rate of interpreting data

Let us look at each of these in turn.

Back to top


‘Order’ is the influence of the whole over the parts for a given number of parts and a given rate. The greater the order, the greater must be the degree of differentiation and the greater must be the limitation of choice.

This means that the higher the order displayed by a model the more limited the number of ways in which a message can be interpreted. If a model displays sufficient order there is only one possible behavioural response compatible with it, and if the variables have been correctly chosen, this response will be the adaptive one, i.e. that making for maximum homeostasis.

Such is the case with the model of basic physical behaviour built by modern physics: each problem has only one possible answer (in terms of the model). Such is unfortunately not the case with sociology where different specialists consider themselves quite free to provide different solutions to the same problem, a situation which they even rationalise as being desirable in order to further the ‘free exchange of ideas’.

Back to top


Complexity must not be confounded with size or number of parts. Information is not simply accumulated. It is organized. It is not the number of parts that is desirable but the variety. The latter term is, in fact, often used instead of complexity. It is not satisfactory, however, because it implies that any variety is useful and this is not so.

Whereas by increasing order one increases the precision with which one can respond to a given environmental situation, by increasing complexity one increases the number of different environmental situations to which an adaptive response is possible.

What is important is that the information permitting these responses must be graduated in accordance with the probability of its being required. A gene-pool, like a brain, does not contain superfluous information or surplus capacity as many scientists affirm. Every item is justified on the basis of the probability that it might conceivably be used, even if it turns out not to be. By reducing this information one is simplifying the informational system and thereby increasing its vulnerability to environmental changes.

Back to top


By increasing the rate of interpretation, one is increasing the rate at which a system can adapt to a changing situation. For this reason, a cultural model as used by human societies and modified by the experience of each generation is more adaptive than a mainly genetically determined one, as used by ant societies for instance.

Back to top

Information value of a message

The information value of a message must clearly be measured in terms of the modification of the model brought about by its interpretation.

If it brought about no such change, it would be irrelevant to the behaviour pattern of the system involved and would simply be ignored. The higher the change, on the other hand, the more relevant it would be, and the greater the effort that the system would make to detect, transduct and organize it.

Consciousness and alertness are clearly nothing more than physiological states favouring the detection of data with high information value.

Back to top

Building up information

How is information built up? Empiricists maintain that information is built up by a process they call ‘induction’, i.e. by observing more and more instances of the same thing. Thus generalities are built up from particularities, but not vice versa. If this is so, why bother to accumulate information? Indeed if one knows nothing about a subject one must proceed by observation. The more one knows about it, however, the more it is possible to predict development without having to depend on observations. Thus one can teach an earthworm to find its way through a maze, but only after a large number of experiences. A rat on the other hand, will learn more quickly, and a man quicker still, because as we ascend the ‘ladder of life’ the nervous system becomes more centralized, and the brain grows so as to contain an even more impressive organization of information. In this way the capacity to deduct particularities from generalities is correspondingly increased.

The same is true as a science progresses. As Professor Elsasser writes,

“the ideal of classical concepts of scientific analysis is to eliminate ultimately all inductive elements and to reduce the description of predictions to pure deduction from general laws, together with the set of parameters characterizing the system in question”.

The sort of deduction we are interested in, however, is not simply deduction from a single general law, but from a hierarchical organization of such laws, i.e. from the model as a whole. Indeed, we believe something to be the case, not because it can be verified empirically, as empiricists would have it, nor because it can be deducted from a general law, but because it is the conclusion that is most consistent with the model as a whole. This must be our most basic epistemological principle.

Back to top

Relationships between the parts

Induction, according to the empiricists allows one to establish cause and effect relationships.

By “cause” they appear to mean the event that ‘triggered-off’ the situation they wish to explain.

Thus if a mother-hen chases a dog away from her chicks, this action must be explained in terms of the stimulus, i.e. the sight of a dog menacing her chicks.

Unfortunately, there are a very large number of possible relationships between the different parts of a model, and a model that takes into account only a single relationship is far too simple to have any predictive value.

Thus the mother-hen chases away the dog so as to ensure the survival of her species, so as to exert a curb on dog populations by limiting their food supply, and because hens have an inborn fear of dogs. All must be regarded as “causes” if the latter term is to have any useful meaning.

To regard only one such relationship as valid, as do Empiricists, is extremely naive, yet it is consistent with the thesis that information can only be built up by induction.

Besides, to use information about the future to build up information about the present or the past – as we are doing in several of these instances – implies goal directedness or teleology, which, as we have seen, Empiricists regard as a major heresy.

Back to top

Connections between the parts

It must be clear that if interpretations and predictions are made on the basis of all the possible relationships that can be established between the different parts of a dynamic or four-dimensional model, then these must be closely connected to each other. It is a truism of ecology that our biosphere is a single integrated system. It can only be so, since it came into being as a single behavioural process, displaying all those qualities that characterize behavioural processes in general.

From this it follows that it can only be represented by a single model, and one whose parts are closely interrelated so that all possible relationships between them can be clearly established.

Back to top

The present divisions of science

At present science is divided into a host of separate disciplines each of which has developed its own terminology and its own method. There is clearly no way of establishing a connection between them, hence the miserable failure of multi-disciplinary research.

These disciplines are only capable of dealing with behavioural processes in laboratory conditions away from the influence of those countless factors that specialists are simply not trained to take into account. To compensate, the predictions they make are taken as being true ‘other things being equal’. Yet there is no reason whatsoever for supposing that they will be equal.

We have seen that the variables left out of a model should be those that are either trivial or extremely unlikely to be affected by environmental change, i.e. that we can establish scientifically as being likely to remain equal. This is a long way from implying as scientists do today that they will remain equal simply because they lack the ability to predict the way they are likely to change, and the effect of such a change on the situation they are studying. If we adopt the former method, the most basic generalities of a process will nearly always be left out, for they are so firmly established that they would be exceedingly unlikely to change. They would, in fact, simply be taken for granted.

Thus, under ideal conditions, a sociologist would take for granted all the information that we possess on the atomic, chemical and biological structure of the societies he was studying and simply concentrate on their cultural organization.

In the same way, a sergeant in dealing with his section will not repeat the orders issued by the army commander to his divisional commanders – he will simply take them for granted.

This does not mean that his men are exempt from these general instructions any more than a society is exempt from the laws governing the behaviour of the atoms, molecules and cells that comprise it.

Unfortunately however, modern sociologists behave as if they were not subject to these laws, not because they have considered the question but simply because they have no means of examining it.

If science were unified, specialized disciplines would continue to exist but since the terminologies used by each one of them would be related to those used by all the others and to that of the unified science as a whole, it would be possible to study them in the light of such general principles; which would only be taken into account if a change in their application could be predicted.

Back to top

Choice of other variables

What is true for the choice of the generalities of the model must be true for that of its other variables.

The different relationships obtaining between the parts of the biosphere that must be taken into account in our model cannot be confined to any spatial or temporal or spatio-temporal sector of it.

To establish these relationships, our model must be able to represent the biosphere as a whole.

Only in this way can the variables be chosen in accordance with their relevance, i.e. their importance and susceptibility to change, rather than arbitrarily, i.e. because they happen to form part of the subject matter of a conventional field of study.

Back to top


It is only in terms of a unified science, capable of describing the biosphere as a whole that it is possible to understand any of its differentiated parts. Such a science would have the following qualities:

  1. Its different parts would be so connected that it would display the necessary degree of order.
  2. It would have the requisite complexity, i.e. would be made up of a sufficiently varied organization of information.
  3. It would permit the correct choice of variables.
  4. It would permit the establishment of that multitude of different ‘causal’ relationships that are required to ensure the correct interpretation of any signal and predict the corresponding change.

The development of such a science would permit scientists to understand the full implications for the biosphere of the specialized work they may be undertaking, and prevent them, if they have any feeling of responsibility to the world they live in, from pursuing their present fatal course.


  • Twitter
  • Facebook
  • Digg
  • Reddit
  • StumbleUpon
  • Diaspora
  • email
  • Add to favorites
Back to top