Notes on: Benton, T.  (1981) 'Realism and Social Science'.  Radical Philosophy, 27, Spring

Dave Harris

The new transcendental realism is suspiciously able to assimilate a variety of positions within Marxism, including Althusser, and Critical Theory, as well as natural science.  The problem is that this is too narrow a view of science.

Bkaskar has written two major books [more since] A Realist Theory of Science (RTS) is about natural science, while The Possibility of Naturalism (PN) is about social science.  Various paramorphisms are established between them.  However the differences between natural and social science are insufficient and are based on the old dualism between the natural and the human which is in turn derived from hermeneutics and neo - kantian philosophy.  This dualism omits some important bits.

In RTS, the main question is what are the conditions of possibility of scientific activities or their intelligibility/rationality?  This includes a discussion of experimentation and scientific education.  However, there are some ambiguities affecting the premisses of transcendental deductions.  For example is the existence of a practice the premiss or its intelligibility or rationality?  Only rationality provides justifiably strong grounds for the ontological conclusions that the world really is like scientific practice.

There were two conditions for the possibility of knowledge, discussed in terms of a transitive and intransitive dimension.  The transitive dimension arises where the object is a cause of knowledge which generates new knowledge; the intransitive arises where the object is a real structure or mechanism, independently of the activities of human beings.  Transcendental deduction operates with the intransitive dimension, since the existence of these independent objects shows that the world has 'ontological depth'.  These can be considered as mechanisms whose powers may or may not be exercised, and may or may not be realized when they are exercised, whose effects may or may not be detectable [see Collier on this].  Effects may be produced in closed or open systems so that scientific laws are necessary abstractions from the range of codeterminations that are possible in the concrete. 

In the transitive dimension, a social ensemble is operated by, but not reducible to, human agency, so that knowledge and beliefs can be seen as antecedents for action.  Social objects are not independent of human beings, unlike the natural ones.  There might be shared characteristics established with natural objects in the world of philosophy, but not in the world of science (14).  There is an additional epistemological obstacle to knowledge in social sciences, because they are unable to close situations experimentally.  Overall, this seems to be a dualism, including methodological dualism between social and natural sciences.  Apparently, the possibility of a naturalistic science of society  and the psyche is discussed, but the whole thing turns on the availability of experimental data.

However, there is an experimental analogue in the human sciences, for Bhaskar, where theories are embedded in practice and are then seen to fail—his example is the failure of neoclassical economics with the onset of depression.  There are problems here, however, in that social science becomes part of social life, not something autonomous enough to be testable by social events.  The analysis also sounds like Popper on social engineering as the test of scientific theories.  Nevertheless, the consideration of the possibility calls into question the rigid boundaries between natural and social science.

In PN, Bhaskar offers some revision of the transitive/intransitive boundary.  Now, social objects can be relatively intransitive, for example, because they can exist and act independently of human knowledge.  There are still problems with the mechanisms that constitute knowledge, however (15) [my notes say that it's not clear why].  Social objects like this can be existentially independent, but they cannot be causally independent.

Can we use transcendental deduction with social sciences too?   The problem is whether there are social science practices in the first place.  Here, the method of transcendental deduction is not clear.  In the first place, it seems to involve a priori deduction of general properties of societies and persons.  However, comparing these properties with those in natural science produces differences.  The demonstration of scientific knowledge is still possible despite or even because of these differences.  However, each proposition has problems. 

Firstly, we are not sure if these general properties are derived from the necessary conditions of social life or from a number of characteristic types of activity, especially with intentional activity, which presupposes social forms according to Bhaskar.  The identification of conditions seems to depend on philosophy after all, and is hardly conclusive.  The whole approach depends on a transformational model of human activity and on controversial definitions, like that of intentional action. It's not sure at this stage if there is still a split between philosophical and scientific arguments.  Above all, Bhaskar seems to be siding with structuralist conceptions of social life, and thus to beg some questions.

Secondly the procedure operates through a number of identifiable limits to naturalism alone.  For example there are ontological limits [to what might be known?] because natural objects are independent of human activity.  However, this is not actually demonstrated, but taken as self evident.  It is controversial when applied to social objects.  For Benton, a power structure exists even if it is not activated, just as the reproductive powers of an organism may never be activated.  There are also ambiguities about human actions, whether they are independent of social formations or vectored by them.  Bhaskar says that social structures do not exist independently of actors' conceptions of them, but this is ambiguous.  For example, some conception of social structures seems essential.  The real issue is the competition between common sense and scientific conceptions, and this competition exists in natural science as well.  Generally, some social formations clearly depend entirely on whether people conceive or recognize them from the point of view of actors, such as a friendship net, but others don't, so that power structures operate with a great variety of conceptions.  Can actor conceptions be seen as causal relations?  They do lead to changes in social structure sometimes, but this is hardly a conclusive demonstration indicating something a priori: instead this opens an empirical question. Bhaskar also says that social objects are less endurable, but this applies to their concrete manifestations. 

Time and space are not in themselves causal for Benton (18), and nor do all laws of natural science correspond to naturalistic mechanisms.  There are differences in the historicity of the social and natural world, however, in that new social facts emerge after theoretical revolutions.  Bhaskar seems to imply a necessary correspondence between social change and conceptual development, as in Marxism, but again this would mean that science is not an autonomous cognitive practice.

Bhaskar notes the difference between open and closed systems, but he overdoes the decisive role of experiments in natural science: this is probably a residual positivism.  Do we need to be able to close a system in order to make a prediction?  Surely we can specify a range of possibilities in open systems?  Methodologically, it is often difficult to isolate the variables in social science, but this is not an epistemological issue which divides natural and social sciences: for example, statistical techniques can be used in social sciences.  Further, not all natural science can be closed—geology and evolutionary biology operate with open systems and also use statistical methods. Bhaskar says there are other limits in that the subject and the object are partially identical in social sciences.  It is still possible that social structures exist independently, however, and conversely, the experiment in natural sciences still requires a subjective intervention. 

There is an implicit status granted to self knowledge, which looks like Lukacs.

Overall, Bhaskar sees the key in the possibilities of experimentation.  Social crises might be useful in revealing the mechanisms of work.  However, such crises also produce new knowledge rather than just testing the old, and even then, only in some societies.  Bhaskar offers no account of the visibility of generative structures especially when crises polarize populations—here, there is no agreement about what is being made visible.

For Benton, the approach features an unnecessary polarization between the natural and the social sciences.  It is over selective when discussing natural science, and over stresses the role of the experiment.  Conversely, it is over dependent of the problematic of the subject in social science.  Ending such selectivity and restoring that which is omitted would confirm the broad outlines of the realist model, however.  It is clear that there are differences between the natural and the social sciences, but these are methodological rather than epistemological and not reducible to a single dichotomy.  Rather, the two display a 'family resemblance'.

more social theory