The Proper Study of Mankind

by Isaiah Berlin

  Let us look at this from another angle. The natural sciences largely consist of logically linked laws about the behaviour of objects in the world. In certain cases these generalisations can be represented in the form of an ideal model – an imaginary entity whose characteristics are by definition what they must be if the entity in question obeys the general laws in question, and can be exhaustively described solely in terms of obeying these laws; that is, it consists of nothing but what instantiates such laws. Such models (or deductive schemata) exhibit most vividly and clearly the laws which we attempt to apply to reality; the objects of the natural world can then be described in terms of the degree of deviation that they exhibit from the ideal model. The degree to which these differences can be systematically described, the simplicity of the models and the range of their application largely determine the success or failure of a given science to perform its task. The electron, the chromosome, the state of perfect competition, the Oedipus complex, the ideal democracy are all such models; they are useful to the degree to which the actual behaviour of real entities in the world can be represented with lesser or greater precision in terms of their deviation from the frictionless behaviour of the perfect model. This is the purpose for which the model is constructed; its usefulness corresponds to the degree to which it fulfils it.

  Such a model or deductive schema is not much in evidence in normal historical writing; if only because the general propositions out of which it must be constructed, and which, if they existed, would require precise formulation, turn out to be virtually impossible to specify. The general concepts that are necessarily employed by historians – notions like State or development or revolution or trend of opinion or economic decline or political power – enter into general propositions of far lesser range or dependability (or specifiability) than those that occur in even the least developed natural science worthy of its title. Such historical generalisations turn out too often to be tautological, or vague or inaccurate; ‘All power tends to corrupt’, ‘Every revolution is followed by a reaction’, ‘Change in the economic structure leads to novel forms of music and painting’ will yield, taken with some specified initial conditions – for example ‘Cromwell had a great deal of power’ or ‘A revolution broke out in Russia in 1917’ or ‘The United States went through a period of radical industrialisation’ – scarcely any reliable historical or sociological deductions. What is lacking here is an interconnected tissue of generalisations which an electronic brain could mechanically apply to a situation mechanically specifiable as relevant. What occurs in historical thinking seems much more like the operation of common sense, where we weave together various prima facie logically independent concepts and general propositions, and bring them to bear on a given situation as best we can. The capacity to do this successfully – the ability to ‘weave together’, ‘bring to bear’ various concepts – is a skill, an empirical knack (sometimes called judgement) which electronic brains cannot be given by their manufacturers.

  At this point we may be told that the mysterious capacity of weighing or assessing a concrete situation, the arts of diagnosis and prognosis (the so-called faculty of judgement) is not unique to history and the other humane studies, or to thinking and decision-making in ordinary life; for in the natural sciences too the capacity for perceiving the relevance of one rather than another theory or concept to the solution of a given problem, and the ‘bringing to bear’ (sometimes with the most dramatic effect) upon a given body of data of notions sometimes derived from very remote fields, is nothing if not the peculiar skill of a gifted investigator, sometimes amounting to the insight of genius, which techniques or machines cannot in principle be made to replace. This is, of course, true; yet there exists one striking difference between the canons of explanation and logical justification used by the sciences and the humanities that will serve to indicate the difference between them. In a developed work of natural science – say a textbook of physics or biology (I do not refer to speculative or impressionistic discourses which are to be found in scientific treatises) – the links between the propositions are, or should be, logically obvious; the propositions follow from each other; that is to say, the conclusions are seen logically to follow from premisses, either with demonstrative certainty, or else with varying degrees of probability which, in the sciences which use statistical methods, should be capable of being estimated with a fair degree of precision. Even if such symbols of inference as ‘because’, or ‘therefore’, or ‘hence’, were omitted, a piece of reasoning in mathematics or physics or any other developed natural science (if it were clearly set out) should be able to exhibit its inner logical structure by the sheer meaning and order of its component propositions. As for the propositions that are stated without argument, these are, or should be, such that, if challenged, their truth or probability could be demonstrated by recognised logical steps from truths established experimentally and accepted by virtually all the relevant specialists. This is very far from being the case in even the best, most convincing, most rigorously argued works of history. No student of the subject can, I think, fail to note the abundance in works of history of such phrases as ‘small wonder if’, ‘it was therefore hardly surprising when’, ‘the inevitable consequences swiftly followed’, ‘events took their inexorable course’, ‘in the circumstances’, ‘from this it was but a short step to’, and most often of all the indispensable, scarcely noticeable and deeply treacherous ‘thus’, ‘whereupon’, ‘finally’ and the like. If these bridges from one set of facts or statements to another were suddenly withdrawn from our textbooks, it is, I think, not too much to say that the transition from one set of statements to the other would become a great deal less smooth: the bald juxtaposition of events or facts would at times be seen to carry no great logical force in itself, and the best constructed cases of some of our best historians (and lawyers) would begin – to minds conditioned by the logical criteria of natural science – to seem less irresistible.

  I do not mean to imply that the humanities, and particularly history, take their readers in by a species of confidence trick – by simulating the outer shell, the logical structure, of scientific method without its substance; only that the force of such convenient, and perhaps indispensable, links as ‘because’ and ‘therefore’ is not identical in the two spheres; each performing its own legitimate – and parallel – functions, and leading to difficulties only if they are regarded as performing logically identical tasks in both spheres. This point will, I hope, become clearer still if it is further developed.

  Let us assume that a historian who is attempting to discover and explain the course of a large historical phenomenon, such as a war or a revolution, is pressed to state those laws and general propositions which alone (at least in theory) could justify his constant use of such logical links as ‘hence’, ‘therefore’, ‘the unavoidable result was’, ‘from this there was no turning back’ and the rest of his stock-in-trade, what could his answer be? He might hesitantly trot out some general maxims about the influence of environment or a particular state of affairs – a bad harvest, or an inflationary spiral, or a wound to national pride – as it affects men in general or a specific group of human beings in particular; or he might speak about the influence of the interests of this or that class or nation, or the effect of religious convictions or social habits or political traditions. But if he is then pressed about the evidence for these generalisations, and upon marshalling what he can, is then told that no self-respecting natural science would tolerate so vague, unsifted and above all exiguous a body of factual evidence, nor such impressionistic methods of surveying it or deriving conclusions from it, he would not (if he were honest or wise) insist on claiming the authority of the methods of a fully-fledged natural science for his activity.

  At this point someone might quite correctly point out to him that not all social sciences are in so deplorable a condition; that, for example, there exist disciplines – economics is perhaps the best known – where something resembling
scientific procedure does appear to take place. In economics concepts can, we are assured, be defined with a fair measure of precision: there is here to be found distinct awareness of the differences between definitions, hypotheses and inductive generalisations; or between the empirical evidence and the conclusions drawn from it; or between the model and the reality to which it is applied; or between the fruit of observation and that of extrapolation; and so forth. This is then held up as a model to the unfortunate historian, wandering helplessly in his dark and pathless wood. Yet if he tries to follow such advice, and to apply to his own subject apparatus recommended by either metaphysical or positivist discoverers of historical patterns, his progress is soon arrested.

  Attempts to provide history with laws have taken two main directions: all-embracing schemata, and division into specialised disciplines. The first has given us the systems of historiosophers, culminating in the vast edifices of Hegel, Spengler, Toynbee and the like, which turn out to be either too general, vague and occasionally tautological to cast new light on anything in particular, or, when the specific findings of the formulae are tested by exact scholars in the relevant fields, to yield implausible results. The second path leads to monographs about selected aspects of human activity – for example, the history of technology, or of a given science or art or craft or social activity. These do indeed, at their best, satisfy some of the criteria of natural scientists, but only at the expense of leaving out the greater part of what is known of the lives of the human beings whose histories are in this way recorded. In the case of a limited field – say the history of coinage in ancient Syracuse – this is, of course, deliberate and desirable as well as unavoidable; my point is that it is only the deliberate limitation of the field that renders it so.

  Any attempt to ‘integrate’ these isolated strands, treated by the special disciplines, into something approaching (as near as we can make it so) a ‘total’ description of human experience – of what, in Aristotle’s words, ‘Alcibiades did and suffered’4 – comes up against an insurmountable obstacle: that the facts to be fitted into the scientific grid and subsumed under the adopted laws or model (even if public criteria for selecting what is important and relevant from what is trivial and peripheral can be found and employed) are too many, too minute, too fleeting, too blurred at the edges. They criss-cross and penetrate each other at many levels simultaneously, and the attempt to prise them apart, as it were, and pin them down, and classify them, and fit them into their specific compartments, turns out to be impracticable. Wherever efforts to pursue this policy have been pressed with real vehemence – by those who were obsessed with the dominant role of some one factor, as Buckle was by that of climate, or Taine by his trinity of the milieu, the moment and the race, or Marxists by that of base and superstructure and the class struggle – they lead to distortions, and the accounts that result, even when they contain illuminating ideas and aperçus, are liable to be rejected as being over-schematised, that is, as exaggerating and omitting too much, as too unfaithful to human life as we know it.

  The fact that this is so seems to me of cardinal importance, and to carry a crucial implication. For one of the central differences between such genuine attempts to apply scientific method to human affairs as are embodied in, say, economics or social psychology, and the analogous attempt to apply it in history proper, is this: scientific procedure is directed in the first place to the construction of an ideal model, with which the portion of the real world to be analysed must, as it were, be matched, so that it can be described and analysed in terms of its deviation from the model. But to construct a useful model will be feasible only when it is possible to abstract a sufficient number of sufficiently stable similarities from the things, facts, events of which the real world – the flow of experience – is composed. Only where such recurrences in the real world are frequent enough, and similar enough to be classifiable as so many deviations from the selfsame model, will the idealised model that is compounded of them – the electron, the gene, the economic man – do its job of making it possible for us to extrapolate from the known to the unknown.

  It follows from this that the greater the number of similarities5 that we are able to collect (and the more dissimilarities we are able to ignore) – that is to say, the more successfully we abstract – the simpler our model will be, the narrower will be the range of characteristics to which it will apply, and the more precisely it will apply to it; and, conversely, the greater the variety of objects to which we want our model to apply, the less we shall be able to exclude, and consequently the more complex the model will become, and the less precisely it will fit the rich diversity of objects which it is meant to summarise, and so the less of a model, of a master key, it will necessarily be. A theory festooned with ad hoc hypotheses to account for each specific deviation from the norm will, like Ptolemy’s epicycles, in the end cease to be useful. Exclusion – neglect of what is beyond the defined frontiers – is entailed in model-building as such. Hence it begins to look as if, given the world as it is,6 the utility of a theory or a model tends to vary directly as the number of cases, and inversely as the number of characteristics, which it succeeds in covering. Consequently one may, at times, be compelled to choose between the rival rewards of increased extension or intension – between the range of a theory and the richness of its content.

  The most rigorous and universal of all models is that of mathematics, because it operates at the level of the highest possible abstraction from natural characteristics. Physics, similarly, ignores deliberately all but the very narrow group of characteristics which material objects possess in common, and its power and scope (and its great triumphs) are directly attributable to its rejection of all but certain selected ubiquitous and recurrent similarities. As we go down the scale, sciences become richer in content and correspondingly less rigorous, less susceptible to quantitative techniques. Economics is a science precisely to the degree to which it can successfully eliminate from consideration those aspects of human activity which are not concerned with production, consumption, exchange, distribution and so on. The attempt to eliminate from the consideration of economists psychological factors, such as, for instance, the springs of human action, or the variety of purposes or states of mind connected with or expressed by them; or to exclude moral or political considerations, such as, for example, the respective values of motives and consequences, or of individual versus group satisfaction – such a procedure is wholly justified so long as its sole aim is to render economics as much of a science as possible: that is to say, an instrument capable of analysis and prediction. If anyone then complains that economics, so conceived, leaves out too much, or fails to solve some of the most fundamental problems of individual and social welfare – among them questions which had originally stimulated this science into existence – one is entitled to reply that the omitted sides of life can be accommodated, and moral, psychological, political, aesthetic, metaphysical questions can perhaps be answered, but only at the price of departing from the rigour and the symmetry – and predictive power – of the models with which economic science operates; that versatility, richness of content, capacity to deal with many categories of problems, adaptability to the complexities of widely varying situations – all this may be purchasable only at the expense of logical simplicity, coherence, economy, width of scope, and, above all, capacity to move from the known to the unknown. These latter characteristics, with which Newtonian physics had, understandably enough, hypnotised the entire intellectual world, can be obtained only by drawing precise frontiers for a given activity and ruthlessly casting out (so far as possible) whatever has not been provided for in this specification. It is for this reason that even in the case of the more descriptive and time-bound (biological and genetic) disciplines, the more general and rigorous the concepts involved and the more ‘technical’ the approach, the better able they are to use methods similar to those of the physical sciences; the more elastic their concepts and the richer their content, the remote
r from a natural science they will be.

  If this is true, then there is a good deal in the Comtean classification of the sciences: mathematics, physics, biology, psychology, sociology are indeed rungs in a descending order of comprehensiveness and precision, and in an ascending order of concreteness and detail. General history – the richest of all human studies – shows this very plainly. If I am purely an economic historian, I can probably establish certain generalisations about the behaviour of some commodity – say wool – in some portion of the Middle Ages, for which enough documentary evidence exists to enable me to establish correlations between the production, sale, distribution of wool, and so forth, and certain other related social and economic facts and events. But I am able to do this only by averting my gaze from questions – sometimes very important and fascinating ones – about other characteristics of the wool-producers or wool-merchants; at least I do not attempt to establish measurable correlations between the sources and movements of the bales of wool and the religious and moral and aesthetic attitudes of wool-growers or wool-users, and their political ideals, and their conduct as husbands or citizens or churchmen, all at once. For the model which attempted to deal with all these aspects of life would (as things are) lose in predictive power and the precision of its results, even if the story gained in comprehensiveness, richness, depth and interest. For this reason, I find it useful to employ technical terms (always symptomatic of the fact that a model is at work) in an artificially delimited field – namely that of economic history. The same considerations apply, for example, to the history of technology, or of mathematics, or of clothing, and the like. I construct the model by abstracting; by noting only what, say, industrial techniques, or mathematical methods, or methods of composing music, have in common, and constructing my model out of these common characteristics, however much of general interest I may be leaving out. The more I wish to put in, the more over-weighted and, in due course, cluttered up and shapeless my model is bound to become, until it is scarcely a model at all, for it no longer covers a sufficient number of actual and possible cases in a sufficient variety of places and times. Its utility as a model will steadily diminish.