Home Articles Reader Opinion Editorial Book Reviews Discussion Writers Guide About TCRecord
transparent 13

Laboratory for Cognitive Studies of Work: A Case Study of the Intellectual Implications of a New Technology

by Laura M. W. Martin & Sylvia Scribner - 1991

Study investigated how new technology affects the working procedures and mental activity of industrial machinists, examining how machinists learn to use computer numerical control technology. Results indicate areas of cognitive difference requiring further study (e.g., differences in conceptualization, formalization, and perspective, and shifts to logical cues from sensory ones).(Source: ERIC)

This article was prepared under the auspices of the National Center for Research in Vocational Education at the University of California, Berkeley. We wish to thank John Antignani, Sebastian DeGiorgio, Dennis Houlihan, Donald Kennedy, Joseph Washington, the members of the Laboratory for Cognitive Studies of Work, and our many advisors in the field for sharing their wisdom with us.

In this article, we review a number of themes that emerged as we undertook a study of how new technology affects the working procedures and mental activity of industrial machinists.

This investigation, which focuses on computer numerical control (CNC) systems, is one of a number of case studies of work activities now being conducted in the Laboratory for Cognitive Studies of Work. For five years we have been analyzing the intellectual demands of a variety of salaried (e.g., production planner, expediter) and hourly jobs at various skill levels (stock room inventory man, machinist). We have selected these particular occupations for study because the introduction of new technologies is rapidly changing their content and requiring workers to master new forms of knowledge and new operational skills. Although each case study focuses on a particular occupation and technology, together they comprise an integrated program of research. The central objective of this program is to develop new conceptual models of workplace learning that will provide a more adequate basis for educational and training programs than current models offer. In the light of this objective, each occupation we investigate serves as a domain for model building as well as the object of descriptive research.

The case study of concern here—investigating how machinists learn to use computer numerical control technology—is of special interest from an educational point of view. The machinists we studied take metal “stock” (usually rods) and, using a variety of machine-mounted tools, carve the raw metal into intricate parts that have holes, threads, slots, tapers, studs, grooves, and so forth. The job is highly technical because the machinists must not only know how to translate the geometry of the blueprints into objects but also must know how to select appropriate tools and adjust and repair the machines that drive them. CNC involves the introduction into machining of multipurpose machines in which manual or cam-driven tool operations are replaced by electronic commands executed through a computer program. Machinists and engineers who traditionally rely on manipulating machine parts to prepare for production runs now must learn how to program, edit, and monitor the electronic control system. Thus CNC technology serves as a prototype of changes in work that require integration of traditional machining knowledge with the symbolic knowledge and logical skills involved in the new “informatics.“1 Our research addresses the question of what form this integration takes, and how it may differ for individuals who already know traditional machining but not programming, those who know programming but not machining, and those (primarily vocational education students) who are learning the new technology “from scratch.” (See figure 1.)

We bring to this case study a theoretical framework and general methodology that was developed in prior research on learning and thinking in the workplace.2

The theoretical perspective stems from the work of L. S. Vygotsky, who first introduced the concept of “mediated activity” into the psychology of thought and language.3 Unlike other approaches to mental functioning, activity theory views cognitive and motivational processes as embedded within larger activity structures whose goals they serve.4 Activity structures involve “mediators”—tools and symbol systems-that have deep implications for the way in which intellectual tasks are accomplished. Thus, this theory suggests that the introduction of new systems and tools into work activities may be expected to change intellectual aspects of these activities. According to the theory, however, the nature of these new intellectual demands cannot simply be projected from a study of the tools themselves. These demands are not all “built into” the tools per se; many of them stem from the way new tools are utilized—the functional purposes they fulfill and the way tasks involving them are socially distributed. A cognitive analysis of the impact of new technologies, therefore, must be concerned with the varieties of ways such technologies are drawn into ongoing activities.

A concrete example helps to illustrate the difference between more standard approaches and the activity approach to cognitive studies. Consider the case of literacy. A writing system is readily recognized as a form of “intellectual tool”5 that mediates a multitude of social practices in our society—educational activities, work activities, recreational activities, and the like. Here is one “technology” that scholars have long agreed has cognitive implications. A long tradition within the scholarly disciplines, and more recently in anthropology and psychology, has sought to derive these implications from a study of the properties of writing (such as the fact that writing objectifies language, is composed of units that are not marked off in speech, and the like).6 This school of thought has put forward claims that these intrinsic properties of writing systems, especially alphabetic scripts, promote abstract and logical thinking among those who master them. Literacy education programs have often been built around this understanding—that literacy both requires and fosters such specialized “higher order” ways of thinking.


Recent empirical work, much of it conducted within the Vygotskian framework, disputes this truism.7 Studies of literacy in various cultural and community settings demonstrate that there is no hard-and-fast relationship between “literacy” and “cognitive implications.” Intellectual implications of literacy are variable, and often contingent on the functions writing serves. If literacy consists only in rote memorization of a sacred text, its intellectual consequences appear limited to specific rote memorization skills.8 Literacy that serves multiple communicative purposes, on the other hand, appears to foster skills in organizing and expressing complex information in instructional situations although it has little impact on memory skills.

The position that introduction of a writing system into a society has a fixed set of cognitive consequences in all places at all times is an argument in the technological determinist vein. In contrast, the activity-theory approach assigns a leading role to the organizational structure of the society adopting writing (is writing the prerogative of a priestly class or available to many social strata?), to the specific practices in which writing is introduced as a mediator (is it confined to private uses or does it figure in trade, government, and everyday life?), to the individuals who are recruited to literacy in the conduct of these practices (do all participants become literate or do some serve as representative scribes?), and to the conditions under which they use it (is text easily produced?). Since in the modern world there is considerable commonality in the functions of literacy across various societies, we might expect such commonality to be reflected in “like” cognitive correlates. It would be misleading, however, to argue backward from discovered similarities in literacy’s consequences to conclude that these are all inherent in the properties of writing systems. The properties of writing systems have certain potential effects on social and psychological processes, but the realization of those effects in turn is dependent on existing, historically created social and psychological factors. The relationship is reciprocal, not one way.

The framework we have illustrated with respect to literacy guides our approach to research on new technological systems such as CNC systems. The general message is that the unit of analysis for cognitive studies of new technologies cannot be restricted to the technology itself, nor to isolated tasks removed from the context of their performance. Such analyses would provide only partial and possibly misleading information for policymakers and practitioners who are concerned with defining educational goals for the future. Rather, research needs to begin with a broader view—an analysis of the societal conditions, institutional settings, and activity structures into which new tools and symbolic systems are being introduced. In the case of manufacturing technologies such as CNC, we need to know how production was carried out before the advent of this technology, its potential for reorganizing these activities, and how this potential is being realized in various sectors of industry. With this knowledge, we are in a better position to identify those changes that may be critical from a cognitive and educational point of view.

In accordance with this outlook, we studied CNC technology, seeking to place it within the context of other technological changes in manufacturing and in the machining industry in particular. We also examined the varied forms in which CNC technology is being implemented, and the forms in which CNC training is now being conducted. We then moved down a level to more detailed studies of CNC technology in use. From these studies, we have generated a set of questions about the new learning demands of this technology that will become the focus of planned learning studies.


CNC is one of a class of new electronic technologies that are fundamentally affecting the traditional organization and definition of many jobs in the industrial workplace. Work sites are undergoing physical transformations as well as changes in job category distribution.9 New social interaction patterns are engendered by the new electronic environments as well.10 Specific performance demands are also shifting, for example, from manual and sensory discrimination toward the processing of textual output,11 suggesting that technology is affecting basic psychological activity.

Such trends raise concerns that are reflected in the literature of a number of fields: management, which is examining workplace communication and control structures;12 manufacturing, which is exploring the design of cost-cutting tools and their impact on shop-floor efficiency;13 labor market analysis, which projects future work-force needs;14 and ergonomics, which seeks to understand efficiency under these new conditions.15 Issues concerning the training of workers for these new environments have also been raised by labor, by industry and government representatives,16 and by leaders in the educational field.17

Although coherent policies are not characteristic of these fields at this time, some common themes are emerging. One theme concerns human factors in the new workplace environments. That is, each particular analytic frame of reference acknowledges that what workers bring to the job—their experience, their knowledge, their familiar modes of operation, their expectations and attitudes—is critical to consider in the design, understanding, and implementation of new technology-based work.


Three lines of research concerned with this theme-namely, the relationship between changing workplace environments and existing worker knowledge—are especially relevant to the study of CNC learning.

One such line is exemplified by multidisciplinary research on work design that is particularly advanced in Europe. In a number of countries, work research institutes and centers have been established, either as free-standing agencies (e.g., the Norwegian Work Research Institute) or as centers housed within academic institutions (e.g., the Swedish Center for Working Life, University of Gotenburg). Combining the perspectives of social scientists, engineers, and psychologists, these research programs seek to understand and preserve the contribution of worker knowledge and skills within modernized work settings. In Norway, for example, occupational protection legislation mandates that employees on the job be given training necessary for them to learn new systems and to take part in planning their implementation. The act further provides that “conditions shall be arranged so that employees are afforded reasonable opportunity for professional and personal development through their work.“18 Fulfillment of these mandates requires an expanded knowledge base dealing with the institutional conditions for learning and the implications of different forms of technological implementation for employee training and development. A variety of research projects have addressed these issues.19 The Swedish Center for Working Life conducts research along similar lines, and academic researchers within the University of Gotenberg direct specific projects, sometimes sponsored by employers and/or unions, that address particular problems of training for changing technologies, Researchers based at the University of Helsinki, Finland, have developed a general methodology for studying working practices in professions and occupations as a means of identifying the major problems posed by changing conditions.20

Much of this research on work design and training issues has reemphasized the importance of taking worker knowledge into account in designing new systems. The research also points to the need for very fine grained analyses of the particular work sites in which a given technology is being introduced, so that the best articulation can be attained between existing and new forms of knowledge and skills.

A second line of research concerns the psychological aspects of technological and organizational changes in work. In this country, research in the arena of industrial psychology has been augmented by research in the cognitive science and artificial intelligence traditions. Cognitive psychologists, who have been interested in human-machine interactions for a number of years, have examined how features of the new environments are differentially interpreted and mastered.21 In such analyses, the content of a task domain is described componentially and a theory of the operator’s conceptual representation of the domain is developed. Mapping the common representations facilitates the design of effective electronic environments. A prominent research strategy has been to compare novice and expert practitioners in a domain with respect to their understandings, modes of problem solving, and the like, and to use this knowledge to design specific training programs. Such studies have been conducted among electronic troubleshooters and employees using other complex devices.22

Despite a sophisticated corpus of work along these lines, cognitive psychologists have paid little attention to unpackaging the more contextually embedded cognitive changes expert workers may undergo in making transitions from one set of technology-based skills to another. This would require attention to such questions as how the new technology is incorporated into on going workplace practices, how its responsibilities are distributed across skill and experience levels, and how different work experience contributes to learning. Some research conducted overseas already suggests that these questions are of critical importance in understanding the learning demands and educational needs of the work force affected by new systems.23

Finally, case studies of new information technologies, although not directly motivated by cognitive concerns, are yielding a set of observations concerning intellectual demands that are directly relevant to our work. These include the observation that there is a growing dependence on manipulating words and symbols that are indirect representations of physical processes,24 that there is a narrowing of the distinction between planning and operative functions,25 and that some functions previously handled by people are taken over by machines while others are not.26

These lines of research suggest that relations among psychological, technological, and organizational changes are complex. Understanding them requires a multilevel analysis at the intersection of cognitive science, other psychological research, and social system analysis.

Scribner has developed such a multilevel approach, which integrates ethnographic description of work activities as they occur in the workplace with quasi-experimental and simulation approaches to the analysis of cognitive processes.27 In keeping with the basic assumptions of activity theory, this multilevel approach views labor as a goal-directed and social activity that promotes psychological functions in the practitioner.28 Various labor activities can be differentiated from one another by the type of tools employed, yet the tools themselves have no fixed meaning; rather, they have an “assigned function.”29 Tools acquire meaning through their use. What is internalized, therefore, is not a simple image of the tool, but the practical experience the operator has with it.30 With a change in tools (e.g., with a transition to electronic tools), a new system of activity comes into being and psychological representations and processes undergo change.

An analysis of changes in thinking and knowledge acquisition due to changes in work tools necessitates an approach that considers the social-historical setting of tool use since the change essentially constitutes a cultural acquisition process. This contrasts with a human-factors or engineering approach to analyzing work task demands, which begins with an analysis of the tool’s assigned function and examines it in relation to the operator’s knowledge state about that tool, without analyzing how the operator acquired that knowledge. The implications of the activity-theory approach for framing educational questions will be illustrated later.

Activity theory provides an approach to what cognitive psychologists today are calling “situated learning,” or learning that is tied into particular activity systems.31 It contrasts with the classical cognitive approach, which takes the relation between a worker and a tool such as a computer at face value, assuming that the worker controls the tool. When a tool is considered an aspect of an activity system, the control is conceived as less unidirectional,32 and a more sensitive estimate of the functional components of the system becomes possible.

We now turn to a consideration of CNC technology and outline some of the system features that contribute to an activity-based analysis of learning.


About 2 million people or 2 percent of the work force in this country are employed in machining, out of approximately 22 million employed in manufacturing as a whole.33 The industrial shop floor is rapidly being transformed by the advent of CNC technology, one of many advanced electronic technologies being refined for more widespread use in both large- and small-scale industry. Although there is no current estimate available, approximately 103,000 computerized machine tools were in place in 1983 (in up to 26 percent of plants for certain types of NCs).34 The consensus of economic and labor forecasters is that these numbers will continue to grow because, as they see it, the future viability of U.S. industry depends on the acquisition of such technology.35


Mechanical means of machining metal date back to Archimedes. Simple mills and threading machines were used throughout the Middle Ages but it was during the eighteenth century that forms of mechanical control began to be developed (by John Wilkenson, Henry Maudsley, Eli Whitney, and others) such that precise borings and interchangeable parts became possible.36 Cam-following machines appeared in the early nineteenth century while systems that contained feedback mechanisms appeared in the early twentieth century.

The principles of machining remained constant despite the introduction of new materials and of new machine designs that allowed refinements in the scale and precision of what could be machined. With these developments, the machinist’s activity, too, remained recognizable. Relatively recent changes in the source of power of the machine and its control mechanisms did, however, affect the machinist’s activity. In the 1950s, under Air Force contract, engineers at M.I.T. developed the first digital control and feedback machine tools, controlled by “information,” which soon became available commercially. These were multi-axis, point-to-point milling and drilling machines and they ran off instructions from a punched tape fed into an electrical “reader.” The reader sent pulses to a series of servomechanisms that translated the electrical signals into mechanical movements.37 Electronic control devices followed shortly thereafter, along with such refinements as continuous path machining capabilities. In the past decade, with the availability of the microchip, free-standing computer numerical control machines became widely used.

The electronic systems of Computer Numerical Control machines (CNCs) have the advantage of being able to handle the large amounts of information necessary to machine complex parts. They have evolved from machines that are hard-wired for particular functions to general-purpose tools that are flexible by virtue of programmable control software.

A brief experiment in record-playback technology, developed at General Electric in the 1940s, devised a machine control system based on skilled workers’ actual motions, which would be translated into “motional” information and stored.38 The system was not fully developed, perhaps because of its dependence on machinists’ skills, but the experiment demonstrates that advanced machining technology could be and was designed to operate according to an existing human model. In light of this alternative, the discontinuity between a manual model of machining and the logical programmable control model of CNC stands out in bold relief.

It has been argued that the mechanization of metal work has historically represented employers’ efforts to “reduce the amount of human skill required in manufacture.“39 In order to understand the impact on cognitive functioning in the workplace, “reducing the amount of human skill required” must be interpreted within its historical setting. For example, one must know the design intent of a particular technical feature as well as its actual implementation in order to begin to understand how that feature may affect an operator’s activity.40


Current management literature discloses a controversy about social forces at work in the development and implementation of new factory technologies such as CNC. Maryellen Kelley describes this argument as one between Taylorists, who believe management concerns for economy and control drive the design of jobs, and post-Taylorists, who think that the new technology itself determines job design.41 Post-Taylorists conclude that the flexibility made possible by programmable systems is best utilized under conditions of expanded skills and responsibilities for operators: “Smart machines are synonymous with distributed control.“42

Case study descriptions of the new technology appearing in manufacturing and machining journals represent the Taylorist view that the point of the technology is to remove control from the shop floor, that is, to “de-skill” workers, who now determine many aspects of production by virtue of the skills they possess.43 For example, one industry handbook states that computer-aided manufacturing is desirable because it “reduces the human component of manufacturing and thereby relieves the process of its most expensive and error-prone ingredient .“44 An industry consultant claims that computer-aided design systems, which are being developed to create CNC programs of graphic designs directly from job-related data entered by the operator, have the potential for increasing “the extent of management contro1.“45

At the same time, many nonindustry analysts argue that human factors must be attended to in implementation plans46 and that companies “should be seen as coalitions with social responsibility,” anticipating the consequences of technology on the quality of work life.47 Technology implementation can be planned in ways that improve potential for worker learning and development, as the European work suggests.48

On the basis of a comparative study among a sample of over one thousand plants of different sizes, Kelley concludes that a multifaceted trend can be seen, with no particular form of social organization of work (and worker control) necessarily tied either to the technology or to the assignment of jobs to workers.49 Historical evidence, other current research, and experiment abroad also suggest that the technology itself determines little per se in the organization of workplace activity.50 Smith and Transfield, for instance, distinguish purely technical views of implementation that equate it with installation from an “exploitation view” that takes implementation to include not only installation but organizational restructuring as well.51 These case studies demonstrate that “technological change is a social process”52 and that the “training culture” in plants must be developed to accommodate the change.53

We conclude from these analyses of recent changes in the organization of work that attributing the effects of technology to physical design alone (for example, its capacity for appropriating skills) is simplistic. Instead, we see social choices on the part of industry, government, and labor constructing a set of plant-conversion and implementation practices that vary along many dimensions.


The drive to introduce CNC technology as rapidly as possible results in some interesting paradoxes that may be resolved only through answers to fundamental educational questions.

CNC technology affects occupational structure and often results in the displacement of many experienced machinists.54 At the same time, in a labor market with a chronic shortage of traditional skilled labor, a new demand has arisen for workers trained to handle machines controlled by computer programs. Vocational schools have been slow to take up these training demands and report that staffs often have difficulty getting the necessary training themselves.55 Problems involved in retraining workers displaced by the new technology have hardly begun to be addressed.56 Unions have not attempted to provide training or retraining programs since most are still in the process of working out general policy positions on new technologies. As a result, employers have been providing much of the necessary training in the form of on-site practice time and vendor instruction.57 Here another paradox arises: Our interviews with machinists and engineers in a variety of industrial and educational settings across the country disclose that, in a number of cases, employers are departing from customary upgrading practices to train younger workers or nonmachinists (engineers, for example) to operate or train others to operate these machines.58 This means that the skills of experienced machinists may not be available in the new informatic work environments. The consequences of further diminishing the traditionally skilled work force through such practices have not been thought out. Moreover, the traditional and successful apprenticeship training model that industry is federally subsidized to expand is disrupted when less skilled machinists are assigned to learn the CNCs. This practice is often justified on the grounds that the new machines now embody many operations of master machinists. However, since the relationship of traditional and displaced skills to the new skills demanded by CNCs has not been identified in any systematic way, it is difficult to assess the impact of these new nonapprentice training models.

In addition to labor-skills availability, productivity concerns drive the development of different models of CNC training. For example, in some settings we have learned about, the entire manufacturing process of a part may be automated, and the attendant work force reduced by as much as 70 percent.59 In those cases, less experienced personnel may be the ones trained. Elsewhere, usually in smaller job shops, the most skilled machinists (not always the supervisor) are assigned to learn the technology and control it. We have heard of plants in which the shop supervisor (still an hourly worker) gains expertise and trains other workers in-house, and plants where the most newly hired are assigned to learn the technology to save money on senior-level salaries.

The consequences of assigning CNC responsibilities to workers with different backgrounds have not been systematically investigated. Williams notes that in Great Britain when CNC programmers were found to “lack metal-cutting experience” this led to “a range of problems in the program preparation, e.g., tools not cutting to programmed dimensions, tools fouling the workpiece. All the programmes would need minor amendments, and at least 10% of the new programmes would ‘smash up.’“60 The situation was remedied by having programmers attend the first runs of a job accompanied by the skilled machinist, who presumably did some training. Conversations with machinists in a number of settings in this country confirm the occurrence of such consequences, but, as other reports suggest, this is not invariably the case. It should be noted that the exact numbers of skilled versus unskilled machinists being assigned to learn CNC programming at their work places is not possible for us to estimate.


Transforming a piece of raw metal into a refined part is a complex process of engineering and manufacturing. The part must be designed and a blueprint of the design must be drawn, which specifies each dimension of the part, the type of metal to be used, the tolerances, and any special instructions (such as plating requirements). After interpreting the blueprint, the setup machinist composes a work plan in which he decides on the order of machining operations for the part and the tools to carry out the chosen operations. In some cases he must make the tools he needs to complete the job, adjusting the plan as he works. In other cases, parts are manufactured in stages on several different machines—mills, lathes, and so forth, although sometimes all operations are carried out at one work station. The various milling and turning operations are controlled by mechanical gauges and levers or by specially fitted cams that delimit the machining parameters. Part dimensions and tolerances, metal properties, and tool use are literally in the hands of the machinists, particularly setup men, whose knowledge of part geometry, metallurgy, output requirements, and tool functioning is extensive.

When the machinist who operates a CNC machine looks at a blueprint, he must also conceptualize a work plan and tool layout. After that, rather than select cams or crank mechanical gauges by hand to the required dimensions, eyeing the sample parts and adjusting his grip, the levers, or the gauges after each operation, he sits at a desk and writes a step-by-step description in a coded language of what the machine must do to the raw stock. The coded instructions are then punched into a computer console and the CNC can be started. When he runs a sample part, he makes a complete part. He measures its correctness and, if adjustments are needed, he goes back to the screen, calls up the program and edits it.

The electronic instructions of the computer program do not mimic the manual movements of the machinist but they express these movements as notations about the machine and its tools in a Cartesian coordinate space. Thus, the essence of CNC machining is not a further elaboration of traditional machining, but appears to be a decisive rupture with it.

Another break between the old and new systems is that CNC machines now perform many of the calculations that a machinist formally carried out himself. For example, the CNC machine can cut a curve given two end points and a radius whereas an old-style machine or numerical control machine required specification of points on the trajectory. However, since it is impractical to build computer-controlled machines that do only one sort of job, like manufacturing a single type of part, CNC machine operators will be expected to maintain a generalized ability to manufacture a variety of parts and to write programs for each unique piece.

How these developments affect the transmission of traditional skills in mathematical calculation is a matter for further study. Observers have noted that while CNC reinforces existing production engineering methods, more collaboration and flexible division of labor arise among engineers, machinists, and programmers-in fact, that a social system is needed to transmit skills.61 Workers may be required to perform new types of calculations in order to program all-purpose tools, and this skill may depend on knowledge that is no longer part of the machining practice itself. As more and more workers join a computerized work environment and traditionally skilled workers retire, it may be that fundamental knowledge will be transmitted solely through classroom instruction, with consequences we cannot yet anticipate.


No single source provides a comprehensive picture of existing training programs in machining, no less newer programs put in place for CNC. Information presented here on CNC training was secured through interviews, on-site observations, and participation in a vendor-sponsored program.

Machine setup skills (as opposed to production skills) were traditionally acquired through apprenticeship programs. In one such program, the 1974-1978  course of study included four days a week of factory work and one day a week of trade school classes. The classes covered blueprint reading, shop theory and mathematics, and machine operation. In the factory, apprentices were paired with a “lead man” who directed their work. They assisted on a wide variety of jobs, going where they were needed according to the factory demands rather than according to a structured training sequence. Generally, apprentices were given low-level tasks to perform. Access to learning critical skills from the other machinists depended on the apprentices’ technical readiness as well as on social dynamics related to their investiture in the subculture of skilled journeymen.

Although such apprenticeship programs are considered to be formal job-training programs, most experts agree that the “real” learning is that which takes place on the shop floor. As they watch more skilled workers and are assigned responsibility for increasingly complex jobs, novice machinists learn the refinements of their trade, the demands of the plant, and the idiosyncrasies of the technology they work with.

With respect to CNC training, we have identified several programs that purport to give complete training in CNC, including both machining skills and computer programming skills. In one vocational high school, students in the second year of machinist training go through a unit on CNC as they approach the end of their training. They work on a small training CNC lathe and learn to make a part they had previously made on manual machines. Basic commands and principles are covered through a combination of hands-on and blackboard instruction.

In the single public community college program in machine tool technology in our area, students in their fourth (and last) semester this year will be able to take a full course on CNC machining. Basic programming ideas will be covered and then students will write between seven and ten programs for parts of increasing complexity. Instruction is scheduled to combine shop and classroom work; however, the school’s two CNC machines are not yet operational and may not be available for the course. The graduates of this program can take jobs as setters, journeymen machinists, operators, and programmers.

We have not located any industry-based program that offers integrated CNC training such as is provided for in the curricula of the high school and college programs described above. It seems to be a common practice, however, for employers to send selected personnel to vendor-conducted schools for brief training in the controller component of CNC. The vendors are the manufacturers of the CNC equipment, whose service representatives offer technical support to customers. Martin participated in one such vendor program, and secured information on the general pattern of vendor-provided training.62

In contrast with vocational education patterns, vendor-provided training often begins in the factory when a CNC machine is purchased. Working with the machinists recruited to learn the technology, the vendor trainer usually begins by programming an actual job, which as often as not is a very complex one, explaining what he is doing to the attendant machinists. He may then have the trainees compose a second program under his supervision or he may just leave the original program for them to use as a model in the future. This introduction may be followed by more intensive vendor-provided training in a classroom setting. The training is focused on the capabilities of the machine and on the controller’s (that is, microprocessor) parameters. It can be organized to cover the list of basic programming commands (about fifty) and include some actual programming exercises. During the course of learning the command possibilities, discussions arise about applied and theoretical topics such as different ways to do threading, the computer’s memory system, and the best ways to do turning behind a “shoulder” (an elevated plane of metal) on the machine.

It seems to be the case that once a plant has a worker who is experienced in programming, that worker becomes responsible for training others, sometimes in on-site classroom settings as well. Of course, many different configurations can be found for selecting trainees. In large plants, programming functions may be centralized and relegated to salaried workers or to a machinist who does not run machines but simply writes the programs. In smaller shops, machinists and their supervisors may learn the new technology; they will also be operating the machines.

From what is known about good training practices, CNC training would require instructors who are not only skilled in the practice of their craft but also skilled in structuring educational interactions. Findings from cognitive psychology suggest that in designing instruction, methods involving verbal explanation, demonstration and guiding, practice, and transfer of activity all need to be balanced,63 and that the level of knowledge and motivation of the workers must also be considered. Yet vocational educators, like other instructors, often work in isolation so that professional development in such directions is problematic. Their training culture, too, must be understood in order to eventually make it responsive to the current needs of industry.

If practical skills and more symbolic skills are as enmeshed as they seem to be here, the strengths of both the classroom and the factory as training environments must be more fully understood in order to better design optimal learning settings.


Within the activity-based approach to cognition, the distinction between practical and scientific knowledge assumes theoretical significance. Vygotsky considered the two systems to be qualitatively different and to have different histories of acquisition.64 Children and adults acquire practical or everyday concepts, as the term implies, through their participation in practical pursuits and ordinary discourse; scientific concepts, on the other hand, are formally transmitted by means of organized instruction typical of school education. In schooled societies, many domains of knowledge represent a “mix” of these two systems that satisfies criteria of functional effectiveness in different activities, for example in the use of practical and school arithmetic in performing store calculations.65

Empirical investigations of everyday concepts and of scientific concepts have increased in recent years, and have demonstrated important differences between the circumstances of learning for the two conceptual systems.66 For one, the use of verbal labels and of concrete examples differs in the two cases. For another, studies show the significance of understanding the differing goals participants bring to tasks for interpreting their learning processes. The intellectual demands of the use of new tools, labels, and symbols are part of the structure of activity; they are not apart from it, that is, a product organized by it.67 In order to develop education and training from this perspective, rather than analyze what local skills are needed and teach to the technology, one assumes that the knowledge and concepts are embedded in an activity system that needs explication. This perspective, of course, implicates social policy as well as training policy, since it assumes that the way particular tool use “figures in human life-activity” contributes to the meaning and use of the tool.68

As we view the machinist’s task from the time a blueprint is put in his hands, the work can be broken down into several phases, each of which seems to raise questions about cognitive functioning pre-CNC and post CNC. Here, we describe the cognitive questions that have arisen at this point in our investigation.


Knowledge of blueprint reading, tool function, operation order, metrics, and arithmetic calculation is fundamental to machining in any form and especially during setup. Setups are most often delegated to an experienced machinist while a novice may only run the machine, which involves fewer, but related, requisite skills. Knowledge of properties of metals under different stress and temperature conditions, complex mathematics, and tool/part geometry are also component skills of setup and machining. Part designers, for example, will frequently omit dimensions on a blueprint. The machinist may have to use trigonometry to determine the missing specification in order to carry out the job. Finally, factoring in the multiple demands of carrying out work activity in a complex social environment (which necessitates considerations of safety, health, timing, etc.) is a skill of traditional machinists that appears essential to the job. If there is pressure to complete a big job in a hurry, machine speed may be increased depending on the metal, the operation, and the size of the part. Many complex factors must be weighed in making decisions about efficient machining.

Before the introduction of CNC, the machining plan was decided on by the setup man. The order of steps and the tools that would be needed were mentally noted. Parts of the plans were adjusted as needed, during the machining process. Experienced machinists would keep track of solutions to particularly tricky machining programs with notations in their own private “black books,” thereby developing a kind of recipe book in their own individual styles. With the CNC system, in contrast, a fixed, written work plan is prepared in which each tool to be used and its identification number are listed, including a drawing of its orientation on the CNC turret, in order of use. These plans are filed for later reference, since the next operator may want to use a tape corresponding to a similar part already machined. He need only modify the program slightly. The private knowledge of the skilled setter becomes a public record with CNC. To that extent, it becomes a more fixed system and it is not clear what becomes of the variability and improvisational skill involved in the older setup systems.

The setup is the first step in what many of our informants describe as a problem-solving activity. In fact, enjoying attacking such problems is often cited as an attribute of a good CNC machinist. What we understand by this is that although the job involves an external notation system (the setup sheet), it becomes a mental one in many respects, and that the “problem” is one that is solved without reference to material events unfolding in the work environment; in short, the machining task becomes an abstraction.


At the same time that we look at traditional skills that enter into any machining situation, we want to look at what might be differences between the skills used in the old and new ways.

What machinists note immediately in working in this medium is that each step in the machining process must be explicitly spelled out in the program, whereas before the activity remained implicitly organized. Thus, programming seems to call for an explication and interpretation process: On one level, it means describing, albeit in code, what the machine and the materials are doing quite precisely—for instance, in specifying the spatial coordinates of the tool nose relative to the work piece. On another level it means stringing commands together semantically, that is, in a way that replicates what an operator would be doing in sequence, while programming for the consequences of the actions as well.

The actual degree of explicitness or interpretation required is very high. As we observed machinists writing their programs, their speech appeared to shift from computer language to machining language and vice versa. For the most part, they described what they were doing in terms that related to the machining process. At other times, they talked “programmer” talk, that is, they described the movements of the machines in terms of the codes. Although we have only just begun to analyze this phenomenon, work among student machinists by other researchers suggests that point of view and the degree to which machining knowledge has become formalized are related.69

Visualization of operations shifts with the CNCs. The tools must be thought of as existing in geometric planes with reference points, according to the scheme of the machine. Planning and explicating the tool movements in a Cartesian coordinate system requires that the machinist learn a new perspective and perhaps results in a translation process from the former hands-on perspective.


Finally, as a part sample is run, the machinist has a chance to check for errors. In this case, troubleshooting can involve either representational or physical corrections and adjustments, that is, corrections to the program or to the machine itself. For example, if an alarm sounds and the machine halts its operation, this could be because a metal chip has jammed the works or because there is a program error. Or, if a machined part has the wrong dimensions, this may be because the program is wrong or because a physical parameter on the machine has been set wrong.

The sensory feedback of the older technologies changes with the CNCs. Although machinists can still rely on their ears and even their noses to some extent to tell if the machine is turning at the correct speed, for instance, many of the cues they would use to detect errors are no longer available. We do not know if the machinist’s sense of the machine tool alters so that he comes to apply his expertise differently or so that he develops an entirely new framework for troubleshooting problems. As in the other transforming processes we have mentioned, we suspect that the new task demands are assimilated into earlier problem-solving strategies and thereby produce new forms of activity. This, however, remains to be demonstrated.

In sum, our preliminary analysis thus suggests the following as areas of cognitive difference to investigate: differences in conceptualization and formalization because of the explicit or verbal knowledge required in tool layout and programming in contrast to the implicit knowledge involved in manual machine operation; changes in approach that arise from machining tasks that become “problems” to solve as such instead of activities in which solutions unfold;70 differences in perspective or point of view in planning machining operations caused by having to visualize machine movement in space; changes in conceptualization because of the rule-based thinking required for programming in contrast to the action-based schemata developed through manual work; and shifts to logical cues from sensory ones as a result of changes in feedback mechanisms.

Hypotheses need to be tested in research that carefully explores the transition from the old to the new technologies among machinists and programmers of differing backgrounds. If such research succeeds in identifying the “critical junctures” between these systems, it will contribute to the development of both curriculum content and pedagogic strategy in CNC and related technologies, bridging the practical and the symbolic.

Cite This Article as: Teachers College Record Volume 92 Number 4, 1991, p. 582-602
https://www.tcrecord.org ID Number: 318, Date Accessed: 1/27/2022 10:10:41 AM

Purchase Reprint Rights for this article or review
Article Tools
Related Articles

Related Discussion
Post a Comment | Read All

About the Author
  • Laura Martin
    City University of New York

  • Sylvia Scribner
    City University of New York

Member Center
In Print
This Month's Issue