Sunday, September 20, 2009

Neurath on sociology

Otto Neurath was one of the central figures in the Vienna Circle in the 1930s and 1940s. And he was the most important figure in the group to consider the social sciences within the "unified sciences" of the twentieth century. As noted in an earlier post, the Vienna Circle set the stage for a powerful tradition of "logical empiricism" as the received view in the philosophy of science in the 1950s and 1960s; and many of these ideas played back into the explicit methodologies of various areas of the sciences as well (physics and behaviorist psychology, for example).

Here I want to focus on the assumptions that some of the Vienna Circle thinkers made about the social sciences. Neurath contributed an article to the International Encyclopedia of Unified Science on the foundations of the social sciences (Foundations of the Social Sciences; 1944), and this is the most extensive commentary on social science methodology that was issued by Vienna Circle thinkers. So it warrants a close reading. Neurath was trained in political economy, and he served as a centralized economic planning official in the German Social Democratic Party in Munich. He was also a key member and leader of the Vienna Circle, and his writings and editorial work for the group were highly influential.

So how does Neurath understand the social sciences? And what does he recommend by way of methodology? Neurath begins by treating the subject matter of sociology empirically rather than definitionally: what do social scientists actually study?
I shall speak of sociological statements as members of one big family of statements which may be found in volumes filled with results of research on the behavior of tribes, on the behavior of customs and languages as they spread through mankind, on the behavior of whole nations, on the growing-up of the fine arts in human societies, on the behavior of human groups and representative individuals (e.g., of artists, priests, statesmen, pirates, peasants, workers, and other people in various societies), on the patterns of cities, on the behavior of markets and administration, and on the various ways of personal life within various societies. (1)
In other words, Neurath is taking the content of a wide range of existing social-science studies of human behavior and society as constituting the domain of "social science" -- rather than beginning with an abstract or theoretical definition of the scope and methods of the social sciences. This approach is consistent with the over-arching Vienna Circle attitude of respect for the content and conduct of the various areas of science as they were currently practiced.

And what does he take to be the scientific goal of such research? He isn't entirely explicit, but here is a preliminary statement:
All the techniques for making well-arranged descriptions, finding correlations, and preparing predictions belong to the field of scientific practice with which I have to do here. ... Up to now we have met an overwhelming number of expressions dealing with social matters. (1-2)
So a fundamental goal of social-science research and thinking, on this approach, is providing a conceptual system -- a vocabulary -- in terms of which to describe social matters. Social scientists then use those concepts to describe the social data that they discover, and they attempt to discover correlations and make predictions based on the statements they have arrived at. Along the lines of Vienna Circle thinking about "criteria of significance," Neurath suggests that sociological concepts need to have clear and specific criteria of application to observable behavior. He refers to this kind of work as "terminological analysis." And the goal, evidently, is to create a uniform and logically specified language for the conduct of social science research and analysis: "Not only the unification of the sociological language is at stake, but a much more comprehensive unification and orchestration, which leads us to a lingua franca of unified science" (2). The resulting concepts need to have "observation content" -- that is, the researcher needs to be able to specify the connection the concept has to the observation of behavior. It is necessary to provide criteria of application based on observation.

But this is where Neurath's criticisms come in. Not all writings in the social or historical sciences conform to this ideal of conceptual clarity and empirical applicability. Neurath offers as a negative example of a social concept, the idea of the "spirit of nations." He suggests that this concept appears to be incapable of being connected to specific observable facts about social behavior; it functions as a speculative postulate of something inherently unobservable; and it should be avoided. "I suggest that it would be better not to discuss these remarks further within logical empiricism, because I see no way of transforming them into physicalist statements" (4). This corresponds to a very basic prescription for sociologists: make sure that the concepts one uses are logically clear and have explicit connections to observation.
We see how perplexing anthropological and historical analyses sometimes appear to be, because the phraseology is anything but consistent. It will not bolster up any argument to add that something belongs to the "mental world" or that there are "motives behind an action" instead of using a simple correlation phraseology. Generalities, such as "factors of social change," even when they are connected with empiricist statements, do not seem to further descriptions or predictions; they often serve as a kind of healing balsam." (17-18)
Several points in this passage warrant comment. First, there is the requirement of "consistency" of language. Here Neurath is reaffirming the goal of arriving at a universal language for social science research. Second is the rejection of mentalistic vocabulary. The point, I believe, is that the scientist cannot provide any observational criteria for applying the mental term to the individual. He/she may be able to provide behavioral criteria for the mental term; but in that case the mental term can be eliminated in terms of a feature of behavior. Instead of: "Berty broke the glass because he was angry," we can replace the sentence with: "Berty broke the glass; Berty displayed a, b, c features of angry-behavior; angry-behavior is commonly associated with things like breaking glasses." This leads us immediately to behaviorism as a methodological requirement for psychology. And the problem with "factors of social change," evidently, is that the phrase is entirely abstract and vacuous -- and therefore does not help with description or prediction. If, on the other hand, the "factor" were specified more concretely -- say, "rising population density as a factor of social change" -- then presumably Neurath would be satisfied. It would then be possible to study a number of societies; evaluate them for population density and social unrest; and arrive at testable statements about the correlation between the two factors.

Consider an example that I think will illustrate Neurath's point here. Suppose we say that "capitalism exists because of a pre-existing spirit of Calvinism." (Neurath begins to consider this example, understanding both concepts in terms of a set of "human attitudes"; p. 16. His use of the example obviously indicates that he is thinking of Max Weber, but he doesn't explicitly mention Weber.) Neurath would require that we be able to answer a series of questions: what are the criteria by which we apply the concept of capitalism to a specific society or group? What are the criteria by which we apply "Calvinism" to a population of people? And what observations and deductive arguments exist that would allow us to assess the truth or falsity of the claim of causation? If we are unable to answer any of these questions, then the statement is not yet an acceptable sociological assertion. Ultimately Neurath argues that we should refrain from making causal claims (21); but if we have clear criteria of application for C and P, then we can arrive at acceptable statements like "All C societies are P societies" or "Some P societies are not C societies". In other words, we can discover observational correlations in the occurrence of C and P across a range of societies or groups.

Neurath gives attention to the topic of "corroborating and supporting hypotheses" (25 ff.) -- what he refers to as "assaying" the statement (19). Here he takes a Duhemian view (Pierre Duhem, The Aim and Structure of Physical Theory), according to which a scientific account of a phenomenon consists of a network of hypotheses that jointly (not singly) have implications for observation. "We look at a network of hypotheses only, and we cannot say from which hypotheses certain difficulties arise." He denies that "crucial experiments" exist that permit the scientist to refute a single hypothesis based on a single experimental outcome. Instead, the system of hypotheses as a whole gains empirical support through positive findings, and loses empirical credibility through negative findings. "I should therefore say we may 'shake' or 'corroborate' the assertion of a hypothesis and finally prefer some hypothesis; but I should not suggest saying that we may 'confirm a sentence more and more,' because even this 'weak' statement deals at least with some 'limit'" (25).

What I do not find in this essay is a clear conception of a "theoretical construct" or theoretical concept -- that is, a scientifically useful concept that does not have criteria of empirical application. Theoretical concepts in physics are thought to be "non-observational" -- that is, there are no direct criteria for applying the concept of a quark to a specific observable thing in a specific time-space location. Instead, theoretical terms are employed in theoretical hypotheses along with "bridge laws" that permit us to relate the theory to the range of observable evidence. The observation-theoretic distinction becomes a crucial one in later analytic philosophy of science; but it doesn't seem to be explicit here in Neurath's essay. (He does refer to the Newtonian concept of gravity; but he doesn't specifically describe the logic of this concept and how it relates to observation; 26.)

Neurath also advocates caution in the search for predictions about social behavior and social processes: "As social scientists, we have to expect gulfs and gaps everywhere, together with unpredictability, incompleteness, and one-sidedness of our arguing, wherever we may start" (27). He reinforces this point with a strikingly modern-sounding point about the non-linearity of social processes: a small deviation at one end can lead to a major deviation at the other end (28).

So the main elements of a logical empiricist philosophy of sociology are here, in the form of a number of prescriptions:
  • logical specification of concepts
  • specification of empirical criteria of application of concepts to observations
  • description of social phenomena in terms of this conceptual system
  • observational assessment of statements
  • discovery of regularities among descriptive statements
  • avoidance of claims of causation
  • disregard of theoretical concepts
And these prescriptions, in turn, seem to lead to a particular model of sociology research -- one that appears to have guided the conduct of quantitative-statistical sociological research for several decades at mid-century. This is a variant of positivist philosophy of science that most strongly encourages general descriptive concepts, strong empirical procedures, and the discovery of statistical regularities among variables. It is what Andrew Abbott refers to as the "variables" paradigm.

Did the Vienna Circle view of the social sciences have any influence on working sociologists in the 1940s and 1950s? Or, more generally, did the writings of logical positivism influence the development of the methodologies and theories of sociology? One direct form of influence can be traced through the priority that behavioral scientists gave to "concept formation" and operationalizability -- for example, Carl Hempel's Fundamentals of Concept Formation in Empirical Science (International Encyclopedia of Unified Science, Vol 2 No 7) (1952) or P. W. Bridgman's The logic of modern physics (1927). The requirement of testability certainly recurs throughout much methodological writing by sociologists in the 1950s and 1960s. The insistence on discovering regularities among social facts plays directly into more sophisticated efforts to understand sociology as a statistical/quantitative science. The idea that the social sciences should avoid using "cause-effect" vocabulary seems to resonate with methodologists who insist that "we discover regularities but cannot assess causal relationships." So there does appear to be a fairly high degree of fit between the views that Neurath advances concerning sociology, and the content of positivist sociological methodology a few decades later.

A final question is this: how should science and philosophy interrelate? Do philosophical findings about scientific method imply that the scientific enterprise needs to start over? In response to this question, the book closes with one of Neurath's most famous passages -- the description of what is now known as "Neurath's raft".
Imagine sailors who, far out at sea, transform the shape of their clumsy vessel from a more circular to a more fishlike one. They make use of some drifting timber, besides the timber of the old structure, to modify the skeleton and the hull of their vessel. But they cannot put the ship in dock in order to start from scratch. During their work they stay on the old structure and deal with heavy gales and thundering waves. In transforming their ship they take care that dangerous leakages do not occur. A new ship grows out of the old one, step by step -- and while they are still building, the sailors may already be thinking of a new structure, and they will not always agree with one another. The whole business will go on in a way we cannot even anticipate today. That is our fate. (47)
This is his metaphor for the reconstruction of the sciences along the lines of logical empiricism and terminological empiricism: gradual reconstruction of the enterprise while underway in the business of performing scientific research.


Vanitas said...

Thanks for the posts on Neurath, which I have read eagerly with a somewhat critical eye. I'm not sure if you're just doing exegesis here, but I can't help but notice that you don't criticize Neurath at all on these matters. This is certainly unusual, for it is almost dogma that the sort of reductionist logical empiricism you're describing here was radically mistaken.

While I don't hold to the established dogma, I certainly see Neurath's conclusions in this post as a reductio ad absurdum against his overall approach. The idea that no psychologist can make a meaningful scientific statement unless it wholly concerns external behaviour is straightforwardly paradoxical. Psychology is the study of the mind, and the exclusion of "mentalese" would annihilate the discipline as we know it. I find the idea that Neurath was truly committed to engaging with scientific practice as it was to be a bit misleading, here, for clearly he would only validate such practice if it conformed to the ideals of logical empiricism.

To sum up an extraordinarily complex point (of which I do not doubt you are aware), there are obviously modes of understanding which are not strictly empirical-causal. The beauty of Weber's observation is not just that it will allow us more precise empirical prediction in the future (though it may do this), but that it facilitates a greater personal understanding of the concepts of protestantism and capitalism, previously taken to be unrelated.

For example, if a protestant-socialist reads Weber and reflects on his observations, she may be confronted with a novel conflict within her personal beliefs, and this conflict may facilitate her intellectual and personal growth. To insist that this sort of end is necessarily unscientific is simply to beg the question against Weber (and most practicing psychotherapists, by the way).

Dan Little said...


Thanks a lot for these good comments. You're right that I looked at the post as exegetical; there are of course very deep criticisms to make about the positions that Neurath advances. I'm trying to slog through a number of the classical positions on the status of the social sciences, and this is one I wanted to look at closely.

I've offered quite a few criticisms of positivist conceptions of the social sciences in earlier postings -- criticisms of the insistence on finding general laws, criticisms of the idea that social science ought to look like the natural sciences, and criticisms of the anti-causal conclusions that Neurath comes to.

I favor a "causal mechanisms" approach to social explanation, joined with a healthy pluralism about the nature of those mechanisms.

I agree as well with your point that the behaviorism that Neurath presupposes is indefensible; I posted on this subject under the category of "mental life" and several other topics.

Vanitas said...

Ah, I see. Thanks Dan, I just started following your blog and perhaps should have looked into past entries.

I think your research is extremely interesting and illuminating, and thanks for sharing it.

Jon Fernquest said...

Thanks for this post. Certainly though inspiring.

I didn't know the logical positivists dealt with sociology. Interesting, even if it was a deadend so to speak.

IMHO Sociology and historiography could borrow more ideas from the philosophy of science.For example, Bayesian philosophy of science provides a perfect critique of the common discourse among historians of trying to determine the one correct reconstruction of history.

With different Bayesian priors one gets different historical reconstructions which allows for multiple "may have happened" histories (in a Bayesian network) instead of one master "actually happened" Rankean history. This is a fruitful way to think of primary source deficient premodern Burmese history.

Ideas like Neurath's may have their principle import as intellectual history that inspires new ideas.

Dan Little said...


Thanks for these comments. I'm curious about how you're thinking about alternative constructions of a narrative based on different "priors". Is it something like this: We say the French Revolution was caused (in part) by a fiscal and governance crisis of the state in 1788-89. We then review conditions in the decade and assign counterfactuals to each of them: this factor could have developed differently, etc. We then try to assign a probability to the occurrence of revolution relative to what we know about 1780. And, perhaps, we judge that "revolution was a significant likelihood but less than 50%." Is it something like this that you're envisioning?

Jon Fernquest said...

The Bayesian idea comes from writing several papers about premodern Burmese history which is information impoverished. Historians end up inserting a lot of speculation which should be made more explicit. Bayesianism could make these assumptions explicit. There is also a style of history that rips the other more speculative history to shreds (Aung-Thwin, Michael Vickery), essentially proving in a way that you can say nothing, which is also kind of unreasonable. Meanwhile none of the historians have a very sophisticated model of how premodern agrarian polities in general worked. This is fairly well studied and lots could be borrowed here as assumptions too. It's really Bayesian to fill out a dirth of information and perhaps provide clues as what sort of evidence is important to look for.