Empiricism, J.S.Mill and Positivism

By | June 27, 2014

You can’t escape theory when doing sociology, or indeed much else. If you think you can or have it’s likely to sneak up behind you and, as the saying goes, ‘bite you on the bum’. But principal among those who often think otherwise are the empiricists and positivists. This blog examines in some detail positivists inclined to believe that data speak our language and do not therefore require translation.

If the discipline of sociology emerged as a product of modernity, it did not do so out of a social or philosophical vacuum. Among its Anglo-American antecedents was a longstanding tradition of empiricism, hinted at initially in the writings of the pre-Socratics but consolidated many centuries later in the foundationalist epistemologies of Locke, Berkeley and Hume. Of these ‘British Empiricists’ Hume is the pivotal figure for us; and most pertinent is tentative account of causality and constant conjunctions in his Treatise of Human Nature, published in 1739/40. It seemed for the phenomenalist Hume that if we are asked for our evidence that A is the cause of B, the only possible answer is that A and B have been constantly conjoined in the past. This does not prove that they are invariably associated. The only recourse then left to us is to admit the psychological origins of our inferences: ‘all our reasonings concerning causes and effects’, Hume insists, ‘are derived from nothing but custom.’ So the connection between A and B consists in the fact that we cannot help – psychologically, not logically – under certain circumstances having certain expectations. If we are asked to justify causal inferences, all we can do is describe how we think.

Hume’s account of causality and constant conjunctions haunts as well as informs positivist approaches, past and present. Comte’s promotion of a positive science of society in his Cours de philosophie positive, issued in six volumes between 1830 and 1842, proved a catalyst. John Stuart Mill, less interested in Comte’s sociology than in the pursuit of a proper methodology for social science, added a refinement that is also a comment on Hume.

For Mill the cause of any event is a set of conditions or factors which, taken together, constitute a sufficient condition for it; his ‘sets of conditions’ replace Hume’s single events. He developed an inductive model of social science from a perspective of uncompromising methodological individualism (leading him to a psychological reductionism inimical to Comte). He spelled out a series of ‘canons’ or procedures of ‘agreement’, ‘difference’, ‘residues’ and ‘concomitant variation’ for testing hypotheses or causal relations. These will strike an immediate chord with disciples of variable analysis.

Two principles underpin Mill’s ‘eliminative induction’: that nothing which was absent when an event occurred could be its cause, and that nothing which was present when an event failed to occur could be its cause. Although Mill did not himself distinguish between necessary and sufficient conditions – in fact, by ‘cause’ he understood sufficient condition not necessary condition – his canons are perhaps best explicated using this dichotomy.

His method of ‘agreement’ states that ‘if two or more instances of the phenomenon under investigation have only one circumstance in common, the circumstance in which alone all the instances agree is the cause (or effect) of the given phenomenon.’ In short, it’s a method intended for determining possible necessary conditions by elimination.

The method of ‘difference’ states that ‘if an instance in which the phenomenon under investigation occurs, and an instance in which it does not occur, have every circumstance in common save one, that one occurring only in the former, the circumstance in which alone the two instances differ is the effect or the cause, or an indispensable part of the cause, of the phenomenon.’ This, then, is a method intended for determining possible sufficient conditions by elimination. Mill also allows for the use of a ‘joint method of agreement and difference.’

The method of ‘residues’ states that if known causes cannot account for a phenomenon it is necessary to seek a cause elsewhere; there must be some residual factor that is not known and/or has not been taken into account. It is a method that relies on deduction rather than induction.

The method of ‘concomitant variation’ states that ‘whatever phenomenon varies in any manner whenever another phenomenon varies in some particular manner is either a cause or an effect of that phenomenon, or is connected with it through some fact of causation.’ It is a method to be deployed when a given factor cannot be removed, rendering the method of difference redundant.

So Mill laid the foundations for positivism anti- and post-Comte. His argument, refined many times over by social statisticians, is arguably more congruent with current positivistic sociological practice than the more ‘sophisticated’ versions or models routinely cited. Paramount among these is the following triad:

  • the deductive-nomological model associated with Hempel, in which premises, statements of general laws and statements of antecedent conditions (the explanans) permit the deduction of a conclusion, a statement describing the event to be explained (the explanandum).
  • the inductive-statistical model, in which the statements of general laws in the deductive-nomological model give way to probabilistic or statistical generalizations, and the relationship between premises and conclusion is one of inductive probability not deductive necessity.
  • the hypothetico-deductive model devised by Popper, and defined here as a variant of (neo-)positivism, in which the emphasis is placed not on (indecisively) confirming a conjecture or theory but on (decisively) falsifying it by a counter-observation.

All of these echo Hume’s regularity theory of causation.

I have previously – in my Health and Social Change – distinguished between three types of positivist investigation routinely conducted under the aegis of the sociology of health and health care. The first of these I called accounting. This refers to the collection and/or collation of data to discern social patterns of behaviour and circumstance, often by means of cross-sectional surveys of large populations.

Secondly, I referred to explaining/predicting. Studies oriented to explaining/predicting aim not merely to identify social patterns of behaviour and circumstance but to explain them; and for the positivist explaining and predicting are necessarily two sides of the same coin. The spirit of Mills lives on in such studies, which tend to follow the inductive-statistical model (Hempel’s model having been largely abandoned and Popper’s largely untried). I will restrict myself here to a couple of critical remarks. Here is Lawson on Hume in his Economics and Reality:

‘if particular knowledge is restricted to atomistic events given in experience, the only possibility for general, including scientific, knowledge is the elaboration of patterns of association of these events. It is thus such constant event patterns, or regularities of the form ‘whenever event X then event Y, that constitute the Humean or positivistic account of causal laws.’

However, as Lawson goes on to argue, the world is not composed merely of experience and events, but also of underlying mechanisms that exist whether or not detected and that govern or facilitate events. This is so for social as well as natural science. Moreover events are typically (a) ‘unsynchonized’ with the mechanisms that govern them, and (b) ‘conjointly determined by various, perhaps countervailing influences so that the governing causes, though necessarily ‘appearing’ through, or in, events can rarely be read straight off’ (Lawson again). In other words, the true theoretical objects of sociological enquiry – that is, mechanisms like class, gender, ethnicity and so on – only manifest themselves in ‘open systems’ where ‘constant event patterns’ or ‘regularities’ of the kind sought by positivists (which exaggerate the potential for experimental ‘closures’) rarely if ever obtain. It must be admitted though that if ‘constant’ event patterns or ‘invariant’ regularities rarely obtain, partial regularities – Lawson calls them ‘demi-regularities’ or ‘demi-regs’ – do. And, as Lawson rightly insists, demi-regs can afford prima facie evidence that relatively enduring and identifiable tendencies are at play. (For an elaboration, see my blogs on basic and dialectical critical realism.)

My third type of positivist investigation was advising. This referred to the prediction of social patterns of behaviour and circumstance with the express purpose of supporting the formation and/or implementation of policy. This is positivism in instrumental guise: it is a pragmatic engagement. Here prediction is key and explanation secondary. I have long maintained that in much clinical or public health research it is prediction that must be prioritized when urgent intervention is required; explanation must needs wait.

So, in conclusion? First, there remain serious philosophical problems with empiricism and positivism. Second, if accounting and advising are legitimate pursuits for sociologists (and I think they are), neither qualifies as a core activity or responsibility for a sociologist committed to explaining social phenomena; and that to my mind is what (professional) sociology is primarily about.



Leave a Reply