Experiment, instrumentation, and procedures of measurement, the body of practices and technologies forming the technical culture of science, have received at most a cameo appearance in most histories. For the history of science is almost always written as the history of theory. Of course, the interpretation of science as dominated by theory was the main pillar of the critique, launched by Kuhn, Quine, Hanson, Feyerabend, and others, of the positivist and logical empiricist traditions in the philosophy of science. Against Carnap, Hempel, Nagel, and Popper, who accorded observation reports an independent status either as a source of inductive support or as a basis for the falsification of scientific theories, Hanson and Kuhn emphasized the theory-ladenness of observation. They made this central point – that all observation is shaped by reference to theory – the cornerstone of a full-blown philosophy of science by buttressing it with two additional lines of argument: First, theories are always underdetermined by the data, several theories being compatible with the same set of data. Hence, choice between theories is never a matter of empirical support, but always turns around conceptual issues. Second, statements derived from theory never confront nature alone; they are always clothed in a web of interrelated beliefs. Thus, theories, their associated observation languages, and the entire technical culture they support must be accepted or rejected as wholes.

This paper is a study of the role of language in scientific activity. It recommends that language be viewed as a community's means of patterning its affairs. Language represents where the boundaries of the community are and who is entitled to speak within it, and it displays the structures of authority in the community. Moreover, language precipitates the community's view of what the world is like, such that linguistic usages can be taken as referring to that world. Thus, language connects on the one side to a community's practical life, and, on the other, to the reality on which it “reports.”

The example treated is Robert Boyle's view of the language proper for experimental practice, and his arguments against the value and legitimacy of mathematical representations within the experimental community. Boyle argued that the use of mathematics was inappropriate because (i) it would restrict the number of potential participants; (ii) it would give rise to unwarranted expectations of the certainty and accuracy to be expected of real physical inquiries; (iii) it would suggest views of natural law and God's relations with the natural world which were incorrect and possibly dangerous; and because (iv) the world upon which the experimentalist operated was not such as was represented by mathematical idealizations. I try to establish that decision about the appropriate language for the experimental community was a moral choice.

Many inventors, engineers, and scientists think in verbal images. Elmer Sperry (1860–1930), a noted American inventor, was able to “operate” in his mind's eye the machines he was developing. For inventors, engineers, and experimental scientists, visualization is often followed by construction of a physical model of the invention, which can be an experimental apparatus. The model, or apparatus, is then tested in increasingly complex environments and changes are made in the physical artifact until it is ready to be used. Examples of this process of development are Sperry's development of a ship stabilizer for the U.S. Navy and a revolving mirror to be used by Albert Michelson in the determination of the speed of light. Thomas Edison called experimentation his development of an invention through the building and testing of a series of models. So, both scientists and inventors experiment. They are not discovering the “secrets of nature”: they are observing how artifacts – their physical creations – behave. These physical models of thought reflect the characteristics of the tools with which they were made. They are socially constructed, as well.

The societal context within which science is pursued generally acts as a productive force in the generation of knowledge. To analyze this action it is helpful to consider particular modes of mediation through which societal concerns are projected into the very local and esoteric concerns of a particular domain of research. One such mode of mediation occurs through material systems. Here I treat two such systems – the steam engine and the electric telegraph – in the natural philosophy of William Thomson (Lord Kelvin).

The steam engine illustrates conceptual mediation. It simultaneously instantiates “labor value” in political economy and “work” in engineering mechanics, thereby identifying the two concepts in the region of their common reference. The partial identification carries with it a structural analogy between a network of concepts from political economy and a similar network in natural philosophy, providing a potent heuristic for the reformulation and further development of dynamics.

The electric telegraph illustrates methodological mediation. It projects the interests of engineering and industry into the interests of electromagnetic theory and vice versa, thereby establishing, in Thomson's view, a Baconian unity of truth and utility. As the common reference of theory and practice, the telegraph locates the truth of theoretical knowledge in its utility and the utility of practical knowledge in its truth.

These particular cases of conceptual and methodological mediation indicate how the local practices, concepts, and interests of a research specialty, or subculture, draw on and are adapted to those of the larger culture within which they develop. Thus the analysis of mediation leads to an ecological model of the social construction of scientific knowledge.

It is often assumed that all sciences travel the path of increasing precision and quantification. It is also assumed that such processes transcend the boundaries of rival scientific disciplines. The history of the personal equation has been cited as an example: the “personal equation” was the name given by astronomers after Bessel to the differences in measured transit times recorded by observers in the same situation. Later in the nineteenth century Wilhelm Wundt used this phenomenon as a type for his experiments on reaction times. For historians of psychology, this has been taken to be an exemplary case where quantified laboratory science rescued astronomy by showing that this was really a psychological phenomenon measurable only in complication experiments. This paper challenges this story. Astronomers neither ignored, nor despaired of, the personality problem. Instead, the managers of the great observatories developed a new chronometric regime of vigilant surveillance of subordinate observers. The astronomers' solution was thus intimately connected with social and material changes in their way of life: a division of labor in the observatories, a network of observing sites, a mechanization of observation. The paper documents these changes and then presents a study of one case where managers, amateurs, and psychologists clashed for authority over the personality problem. Measurement is given its meaning when situated in specific contexts of styles of work and institutions. Disciplines give meanings to values, and often resist attempts by others to redefine these meanings or to gain authority over measurement. Quantification is not a self-evident nor inevitable process in science's history, but possesses a remarkable cultural history of its own.

This paper analyzes the complex and many-layered interrelation between the realization of the inevitable limits of precision in the experimental domain, the emerging quantum theory, and empirically oriented philosophy in the years 1925–1935. In contrast to the usual historical presentation of Heisenberg's uncertainty principle as a purely theoretical achievement, this work discloses the experimental roots of Heisenberg's contribution. In addition, this paper argues that the positivistic philosophy of elimination of unobservables was not used as a guiding principle in the emergence of the new quantum theory, but rather mostly as a post facto justification. The case of P. W. Bridgman, analyzed in this paper, demonstrates how inconclusive operationalistic arguments are, when used as a possible heuristic aid for future discoveries. A large part of this paper is devoted to the evolution of Bridgman's views, and his skeptical reassessment of operationalism and of the very notion of scientific truth.

Idealizations and approximations are an indispensable tool for the scientist. This paper argues that idealizations and approximations are equally indispensable for the philosopher of science. In particular, it is shown that the deductive model of scientific theories is an idealization in precisely the same sense that frictionless motion is an idealization in mechanics. By its very nature, an idealization cannot be criticized as not being absolutely true to the facts, for it need not be. Thus, the usual type of criticism levelled against the deductive model is shown to be irrelevant. The main step in the argument consists in using a logic of approximation to clarify the notions of approximate implication and approximate entailment, which are usually rejected as hopelessly vague. More generally, the conception underlying the paper is that the desiderata of science should be formulated in such a way that complying with them could be a matter of degree, and that pragmatic considerations could be taken into account. This conception is motivated by the wish to introduce some quantitative concepts in the philosophy of science, thereby bringing the philosopher's perspective closer to that of the scientist.

When we actually perform an experiment, we do many different things simultaneously – some belonging to the realm of theory, some to the realms of methodology and technique; however, a great deal of what happens is expressible in terms of socially determined images of knowledge or in terms of concepts of reflectivity – second-order concepts – namely thoughts about thoughts.

The emergence of experiment as a second-order concept in late antiquity exemplifies the historical development of second-order concepts; it is shown to be rooted in the Sophists' cunning reason (Greek metis) and is followed up in the work of Ptolemy, Copernicus and Galileo.

Then, by way of epistemological explication, the three levels of representation of an experiment are shown to be analogous to Baxandall's three levels of representation of a picture.

Finally it is shown that such an interpretation only makes sense in terms of two-tier thinking: realism, inside a conceptual framework which is chosen or arrived at, relativistically.

Behind the dispute over the relative priority of theory and experiment lie conflicting philosophical images of the nature of scientific inquiry. One crucial image arose in the 1920s, when the logical positivists agitated for a “unity of science” that would ground all meaningful scientific activity on an observational foundation. Their goals and rhetoric dovetailed with the larger movements of architectural, literary, and philosophical modernism. Historians of science followed the positivists by tracking experimental science as the basis for scientific progress. After World War II, historians and philosophers of science created an antipositivist movement, inverting the positivist idea that observation had epistemic (and historical) priority. But this counter-movement retained the modernist aspiration to unity, now chaining observation to theory. Once again historians of science, following their philosophical colleagues, illustrated the new modernism with historical instances of observation dominated by theory.

Either reductionist scheme, by privileging one activity over the other, dictates an overly constrictive periodization. We need a wider class of periodization models (“central metaphors”) that will allow instrumentation, experimentation, and theory a partial autonomy without granting any one the sole legitimate narrative standpoint. Such a heterogeneous representation of historical traditions may, surprisingly, make better sense of the perceived coherence of activity across theoretical transitions than the monolithic modernist representation of science it displaces.