Walter A. Shewhart

Shewhart, Walter Andrew (born the March 18th 1891 with New Canton, Illinois, E. - U.; deceased the March 11th 1967 with Troy Hills, New Jersey, the USA); married on August 4th, 1914 with Edna Binder

Higher learning at the University of Illinois, then PhD of physics of the University of California (1917). Assistant of physics to U. of Illinois (1916-17), then chief of the department of physics of Normal Wisconsin School with Lacrosse (1917-18).

Engaged in 1918 with the Western Electric Company, transferred into 1925 to the Beautiful Telephone Laboratories, of which it will be paid until its retirement in 1956. Distinctions: Institute off Mathematical Statistics: founding member and fellow, president (1937, 1944). -- American Society for Quality Control: founding member (1947), first Shewhart Medalist (1948). -- American Statistical Association: president (1945). -- American Society off Mechanical Engineers: Holley Medal (1954). - Member of: International Statistical Institute, Royal Statistical Society (honorary fellow, 1954), and others still.

The name of Shewhart is associated with the methods of Statistical quality control in the industrial productions of great series (the “Shewhart control charts”), but its contribution in the field them technical statistics must be replaced within the framework of a theory of the knowledge which it worked out for the specific context of the activity of industrial production. Shewhart was an original epistemologist, deeply marked by philosophy pragmatist resulting from C.S. Peirce, W. James, C.I. Lewis, which it truly implemented operationally in the field of engineering. He at the same time sought to think and conceptualize the collective aspect of work in company, which is worth to him to be also known in the field of management, particularly thanks to W.E. Deming which was made the promoter of its designs. By its intermediary, Shewhart is well-known in Japan where it has thus, indirectly, played a big role in the development of the methods of management of quality (Kaoru Ishikawa).

Work on the control charts originates in a political will of the direction of the Department of engineering of the Electric Western to define scientifically a step of reflecting quality assurance “has Company-wide view” (1922). Under the direction of R.L. Jones and G.D. Edwards, it is a whole team which attacks these problems soon: for Shewhart, it is necessary to quote H.F. Dodge, E.C. Molina, T.C. Fry, H.G. Romig. When the Beautiful Laboratories Telephone are juridically made up, in 1925, the engineering of the inspection its place in the flow chart has there. It is into 1931 that Shewhart publishes the whole of its work on the control charts, in a masterly treaty which still remains of topicality. The development of these statistical methods of quality assurance is a subject whose Beautiful Labs are shown still proud (Fagen 1975).

Shewhart thus had an institutional legitimacy and average the considerable ones which enabled him to intervene in many places of the company and, in particular, to test on the production sites with real data. Dodge and Romig profited from the same favorable conditions, which led to realistic methods of control of reception per sampling. It is there an essential difference compared to other work on the same subject starting to appear in Europe about this time (France, Germany, the United Kingdom), which remained of limited scale and was finally supplanted, after the Second world war, by the methods of Beautiful Labs.

The Méthode of the control charts is particularly well adapted to a use in the workshops thanks to a clever and elegant exploitation of the charts, which makes it possible to limit the recourse to the mathematical theory. The quality of the products being defined by a whole of measurable physical sizes, supposed to follow statistical distributions, the control chart provides, by a rather simple procedure ultimately, a visual representation of the evolution of the parameters of the distribution (average and dispersion). Shewhart, in all its publications, is expert in the use of the graphic resources (what one can connect to his attraction for the designs semiotics of Peirce). In the plan of the statistical techniques, Shewhart uses initially the theory of the distributions of K. Pearson, which it supplements by work on the small samples of `Student' and Fisher. It introduces the rational concept of “sub-group” to define préocédure and the sampling rate.

But the Control chart is only the emerged part of the iceberg. Even working in industry, Shewhart remained always a physicist, worried by the problems of measurement in experimental physics and by the nature of the process of acquisition of knowledge. He believes deeply in the statistical nature of all the physical laws and tackles face the deterministic designs: We ordinarily think off the physical and engineering sciences exact ace being. In majority off physical measurements this is practically true has. (...) With the introduction off the molecular theory and the theory off quanta, it has been necessay to modify nap off our older designs. Thus, more and more we are led to consider the problem off measuring any physical quantity probable ace that off establishing its most been worth. Conceive We are led to off the physico-chemical laws ace has off statistical determinism to which the “law broad numbers” imparts the appearance off precision infinity. (1924, 1931)

Theoretically to found the use of the control charts, Shewhart works out, by an astonishing work of abstraction, the concept of “constant system off chance causes”. Laplace had already evoked this idea in its analytical Théorie of the probabilities (1815), filiation that Shewhart recognizes. Intuitively, the concept is simple: they are random systems controlled by a statistical distribution not depending on time. The control chart makes it possible to locate the intervention of significant causes of disturbance (called “assignable”), which can then be required and eliminated. Gradually, one thus manages (the experiment proves it, says Shewhart) to eliminate all the assignable causes and the system is then known as “under statistical control”. It also introduces the economic point of view by cost-benefit considerations on this search for causes.

From there, an ordinary engineer could have been satisfied with a pragmatic approach or calculative, but Shewhart makes of it the heart of a theory of production of knowledge on the physical world, in particular in its book of 1939. It shows that there does not exist satisfactory criterion to establish if system is under statistical control; that can be only one assumption subjected to future experimentation. In philosophy pragmatist, knowledge is never sure, it only enables us to make forecasts that the future is likely to cancel. However series of measure to determine a physical size have direction only if the system is under statistical control. Shewhart proposes a ternary design thus (as the sign of Peirce) of the process of measurement in physics and, more generally, production of knowledge: the data provide “testimony”, on the basis of which we can make certain “forecast” together with a “degree of confidence”. In particular, it results from it that a physical measurement is never a result but only one stage in a process of measurement which is spread ad infinitum towards the future, following the example " semiosis" of Peirce. Most important, they are then the conditions installation to bring a stabilization of the process of measurement, which is the work of a human collective. To connect the world of the concepts to that of the human action, Shewhart takes again with C.I. Lewis and to the physicist P. Bridgman the fundamental idea of operationnalisation: “The concept is synonymous with the corresponding set off operations” (Bridgman).

The theories of Shewhart thus throw a bridge between the epistemology of measurement and the design of the human organization which must make function the process of measurement. For him, the process of statistical control is homologous with the process of the scientific knowledge. Control comprises three phases (" specification, production, inspection" , in English in the text) who corresponds respectively to: to conceive an assumption, to undertake an experiment, to test the assumption. “The three steps constitute has dynamic scientific process off acquiring scientific knowledge” (1939, p. 45). These three stages were reformulated by Deming under the name of “cycle of Shewhart” (called " later; coil of Deming") and they intervene in the recent developments of the theory of the organizations such as the organisational training, whose Shewhart seems a precursor thus: “farmhouse viewed production in this way constitutes has continuing and coil-corrective method for making the most efficient uses raw and fabricated material off.” (ibid)

There remains a paradox in the social character of Shewhart. Its writings, truffles of references to physics and very diverse technical fields, are of a difficult reading. According to testimonys, its talks were rather little included/understood industrialists. And he is however the inventor of a technique which has accepted very well in industry, recognized easily usable. Obviously, the weight of AT&T in American industry played a crucial role to promote its work, as from the moment when they were supported by the company. But it is also necessary to underline, according to testimonys, a great personal skill to engage of alliances and to weave relations, institutional relays. He fully played the game of the agencies of standardization, of the learned societies and associations of engineers: since 1929, it obtained subsidies to create committee joint of American Society for Testing Materials and powerful American Society off Mechanical Engineers, which leads in 1935 to the publication of a standard which will be taken again then during the war, for the military market needs. It is also necessary to mention the promotional effect of the conferences which it gave to London in 1932, on the invitation of E.S. Pearson. The British, not wanting to remain behind in this field, undertook work whose several publications testify to Royal Statistical Society, where was created a section of the industrial and agricultural applications. Shewhart was however a modest and not involved man. He believed in the fruitfulness of collective work, of the exchange between disciplines and professions: he saw the role of the statistician in the company like that of the coordinator of the many experts intervening in the production process of knowledge.

Bibliographical references

Books of W.A. Shewhart

Shewhart, W.A., Economic Control off Quality off Manufactured Products . New York: Van Nostrand, London: MacMillan, 1931,501 p. Réédité in 1980 by American Society for Quality. (" Bases of the control of the qualité" , Economica, 1999, ISBN: 271781714X)

Shewhart, W.A., Statistical Method from the Viewpoint off Quality Control . The Graduate School, U.S. Department off Agriculture, Washington, 1939,155 p. Réédité by Dover Publications, 1986 (with has new foreword by William Edwards Deming)

W.A. Shewart in addition published many articles.

Books on W.A. Shewart

Bayart, D.: “Walter Andrew Shewhart, Economic Control off Quality off Manufactured Product”. in Grattan-Guinness, I. (ED): Landmark Writings in Western Mathematics 1640-1940. Elsevier B.V., Amsterdam, 2005,926-935

Deming, W.E., “Shewhart, Walter A.” in: Kruskal, W. (ED): International Encyclopedia off Statistics, New York, 1978, vol. 2,942-944

Eisenhardt, C., “Shewhart, Walter Andrew” in: Dictionary off Scientific Biographies, vol. 18, Supp. II, 816-819 -- with a detailed bibliography

Industrial Quality Control , vol. 24 (1967), special issue in memory of Shewhart -- with many testimonys.

Fagen, M.D., (ED), 1975: In History off Engineering and Science in the Beautiful System, The Early Years (1875? 1925), vol. 1, Beautiful Laboratories Telephone. Chapter 9: “Quality Insurance”.

Random links: