1 Introduction.

Since the first Industrial Revolution, technology has steadily transformed living standards and daily life. The aggregate effects of new technology--rising productivity and improving product performance--are visible effects of from new knowledge of "how to do things." But what is the nature of this knowledge, and how does it evolve over time? This paper investigates long-term technological change and the evolution of enabling knowledge through the lens of a single industry over more than 200 years.

Changes in technological knowledge are usually observed indirectly, as changes in methods or performance. Performance that improves by more than can be explained by measured inputs is taken as evidence of changes in the stock of knowledge. Implicitly this assumes a causal chain approximately as follows: learning activities create new knowledge that allows the firm to implement superior designs and methods that improve local physical performance such as machine speed and material consumption, which ultimately causes better high level performance (Figure 1.1). But generally, the middle variables in this chain are not observed directly.

[FIGURE 1.1 OMITTED]

Our focus is on the intermediate steps of this chain--new knowledge, superior methods, and improved performance at workstations--that cause improved aggregate performance. Changes in production methods are explored in the companion paper, From Filing and Fitting to Flexible Manufacturing: The Evolution of Process Control by R. Jaikumar [15]. Here we explicitly examine the new knowledge that made possible these changes.

Our case study centers on the manufacturing methods of a single company over 500 years. The company, Beretta, has remained in family hands and has made firearms since its founding in 1492, when firearms were manufactured as a small-scale craft with only hand tools. Jaikumar identified six distinct epochs of manufacturing, characterized by different conceptions of work, different key problems, and different organizations (Table 1.1).

Each epoch constituted an intellectual watershed in how manufacturing and its key activities were viewed. Each required introducing a new system of manufacture. Machines, the nature of work, and factory organization all had to change in concert. Within Beretta, each of these epochal shifts took about ten years to assimilate.

A longitudinal study of a single industry is an excellent test-bed to examine technological change over a long period. In Jaikumar's study, the fundamental product concept changed little from the 16th to the late 20th century: a chemical explosion propels a small metal object through a hollow metal cylinder at high speed. With such product stability, changes in manufacturing stand out even more.

The central problem in manufacturing over the entire period was to increase process control, for once society moved beyond making unique items by hand predictability, consistency, and speed were achieved by progressively tightening control. Each new epoch revolved around solving a new process control challenge, generally reducing a novel class of variation. To accomplish this required major, often unexpected, shifts in many aspects of manufacturing (Table 1.2). The nature and organization of work changed, use and sophistication of machines increased, and, most important for our purposes, manufacturing control shifted, all requiring changes in knowledge.

We will describe shifts in technology using the metaphor of transformation from art to science. Jaikumar observed that "The holy grail of a manufacturing science begun in the early 1800s and carried on with religious fervor by Taylor in early 1900s is, with the dawning of the twenty-first century, finally within grasp." (1) But precisely what does this mean? Is such evolution inevitable? Is it universal, or limited to manufacturing?

As late as the early 18th century, making firearms still relied entirely workers' expertise. Documented or standardized methods were non-existent.

Production involved the master, the model, and a set of calipers.

If there were drawings, they indicated only rough proportions and

functions of components. Masters and millwrights, being keenly

aware of the function of the product, oriented their work towards

proper fit and intended functionality. Fit among components was

important, and the master was the arbiter of fit. Apprentices

learned from masters the craft of using tools. Control was a

developed skill situated in the eyes and hands of the millwright.

Inasmuch as adaptive skills are really contingent responses

to a wide variety of work conditions, procedures cannot readily

be transferred. Critical knowledge was mainly tacit, and a

journeyman had to learn by observing the master's idiosyncratic

behaviors. The master, who could solve the most difficult of

problems, fashioned each product such that quality was inherent in

its fit, finish, and functionality. [15, Section 2]

This description corresponds to technology as an art. Learning was by apprenticeship; quality was achieved by rework; progress occurred slowly by trial and error; techniques and knowledge were idiosyncratic.

In contrast, in the most advanced flexible manufacturing systems of the late 20th century people are normally absent from the production area, and machines execute complex contingent procedures under computer control. Operators manipulate symbols on workstations, and use scientific methods of observation, experimentation, and data analysis. Alternative production methods can be precisely described, tested, and embodied in software. Methods and general knowledge can be transferred to other locations, machines, and products with little effort and no face-to-face communication. This is manufacturing as a science. Manufacturing changed profoundly over the two century transition from art to science, with performance improvements on some dimensions of two orders of magnitude or more (Figure 1.2).

[FIGURE 1.2 OMITTED]

Transitions from art toward science can be seen in many technologies. Early aviation, literally a "seat of the pants" technology early in its development, today includes the Global Hawk aircraft, which can take off, cross the Pacific, and land without human intervention. In contrast, although product development technology has progressed tremendously, it still has remains in many ways more like art than a science.

Although we are concerned here with a relatively small industry that has not been leading edge since the mid-19th century, the evolution of knowledge and the transition from art to science are still critical in all high-tech industries, and influence many contemporary issues such as offshoring, automation, and outsourcing. These activities require transfers of knowledge and information across organizational and firm boundaries. We will see that the difficulty of such transfers depends on the detailed structure of knowledge. [18]

In Section 1.1 we consider different ways of classifying technology along a spectrum from art to science. Section 1.2 presents a formal model of technological knowledge that supports precise descriptions of changes in knowledge when learning occurs. Prior research is presented in Section 1.3. The case study evidence is presented in Section 2 and Section 3.

In Section 2 we examine the first three epochs of manufacturing (approximately the 19th century), during which workers' discretion and insight were progressively reduced, culminating in Taylor's extreme division of labor and separation of intellectual work from line operations. We will see that the de-skilling of workers in the Taylor System rested on an unprecedented level of technological knowledge, developed by Taylor himself using several seminal concepts.

In Section 3 we examine the development of knowledge over the last three epochs, in which workers increasingly became problem solvers and knowledge creators, effectively reversing Taylor's de-skilling paradigm. We also examine the integration of formal science with practical engineering. Finally, we consider what happens when novel and immature physical processes are substituted for mature ones. Even when the core physical process is entirely changed, considerable knowledge from old processes is still relevant.

In the concluding section we examine broad patterns of change in manufacturing over the centuries.

1.1. Art and Science in Technology

The metaphor of art and science in human endeavor is long established and widely used. Military treatises speak of the "art and science of war" as in a 1745 book that provides "a short introduction to the art of fortification, containing draughts and explanations of the principal works in military architecture, and the machines and utensils necessary either in attacks or defenses: also a military dictionary ... explaining all the technical terms in the science of war" [3]. Sometimes a clear distinction is made between "art" and "science," as in the title of an American book on surveying circa 1802: Art without science, or, The art of surveying: unshackled with the terms and science of mathematics, designed for farmers' boys [33]. The two are not as clearly differentiated in a 1671 title, An introduction to the art of logick: composed for ... [those who do not speak Latin but] desire to be instructed in this liberal science [28].