The late scientist-philosopher Michael Polanyi wrote extensively of the central place in the general scheme of human knowledge occu pied by knowledge that cannot be articulated- tacit knowledge. On the simple observation “We know more than we can tell,” Polanyi built an entire philosophical system (Polanyi, 1967, p. 4) . Though the full import of Htacit knowing” in Polanyi’s philosophy can only be hinted at by examples of what would ordinarily be called “skills,” such examples do provide familiar and compelling illustrations of phenomena of broad significance. In fact, in Polanyi’s Pe rsonal Knowledge (1962) , the discussion of skills (ch. 4) plays a role analogous to our own discussion here. It provides a useful perspective on other realms of knowledge -in his case, that of scientific knowledge; in ours, that of organizational capability.
To be able to do something, and at the same time be unable to ex plain how it is done, is more than a logical possibility- it is a common situation. Polanyi offers a good example early in his discus-sion of skills: “I shall take as my clue for this investigation the well-known fact that the aim of a skillful performance is achieved by the o bservance of a set of rules which a re not known as such to the person fol lowing them. For example, the decisive factor by which the swimmer keeps himself afloat is the manner by which. he regulates his respira tion; he keeps his bouyancy at an increased level by refraining from emptying his lungs when breathing out and by inflating them more than usual when breathing in; yet this is not generally known to swimmers” (Polanyi, 1962, p. 49) .
The difficulty of explaining the basis of a skilled performance comes to the fore in the teaching or learning of skills. Polanyi’s swimming example suggests that in some cases the difficulty may arise from the fact that the “instructor” is quite unaware of the key principles, and that he actually serves less to instruct than to detect and reward randomly occurring improvements in performance. In other cases, the instructor may be able, or at least be subjectively confident that he is able, to explain the matter in detail. But the de tailed instruction offered typically consists of a list of subskills to be executed in sequence, and the instructions neither convey the ability to perform the subskills with requisite efficiency nor assure the smooth integration of those subskills into the main skill . This point is emphasized by Miller, Galanter, and Pribam, commenting on a description of how to land an airplane: “When skillfully elaborated and executed it will serve to get pilot and craft safely back to earth. It is a short paragraph and could be memorized in a few minutes, but it is doubtful whether the person who memorized it could land a plane, even under ideal weather conditions . In fact, it seems likely that someone could learn all the individual acts that are required in order to execute the Plan, and still be unable to land successfully. The sepa rate motions, the separate parts of the Plan, must be fused together to form a skilled performance. Given the description of what he is sup posed to do, the student still faces the major task of learning how to do it” (Miller, Galanter, and Pribam, 1960, pp. 82-83) ,
Instruction in a skill typically consists in large part of the imposition of a discipline of practice, a portion of which is supervised by the instructor. Verbal instruction is included, but is predominantly in the form of critique of practice. Illustration by the instructor and (attempted) imitation by the learner is often employed as an alterna tive mode to verbal instruction and critique. As Miller et a1 . indicate, verbal instruction by itself-the information in the “how- to-do-it” book-provides only a starting point at best for the acquisition of the skill. Possession of such a book-the articulable portion of the knowledge involved- may be indicative of ambition to learn, but it certainly does not certify possession of the skill.
The limitations of verbal instruction are even more apparent when the learner is attempting to reacquire a skill that has become rusty. Only in extreme cases does the how-to-do-it book prove useful in the reacquisition of a rusty skill. The remnant of the skill itself, lying la tent in the brain, is typically more helpful as a restarting point than any collection of more words could be. What is needed is renewed practice and constructive criticism, not the beginner’s handbook.
These propositions do not relate only to psychomotor skills. With minor modification, they extend to the realm of specific cognitive skills such as facility in mathematical manipulation of a particular type, the ability to solve the theoretical exercises characteristic of a certain area and method of scientific inquiry, or the ability to gener ate good solutions to complex production scheduling problems. The manipulation of equations in elementary algebra will serve as an ex ample . Clearly, the axioms of the real number system together with a relatively short list of problem- solving heuristics (like “isolate the unknown”) do constitute, in a sense, an articulated account of the skill involved. Equally clearly, the skilled manipulator in action has little or no conscious awareness of this articulated characterization of his activity. He does not think “distributive law- rearrange terms – factor out X” and so on, but simply “perceives” productive transformations of the expression and carries them out, often making several transformations at once in the course of rewriting the expres sion. There is, in Polanyi’s terms, only “subsidiary awareness” of the rules being employed, whereas there is “focal awareness” of the expression manipulated.
It seems clear that the “tacitness” of a skill, or rather of the knowl edge underlying a skill, is a matter of degree. Words are probably a more effective vehicle for co mmunicating the skills of elementary algebra than for those of carpentry, and more effective for carpentry than for gymnastic stunts. Also, a trait that distinguishes a good in structor is the ability to discover introspectively, and then articulate for the student, much of the knowledge that ordinarily remains tacit.
The same knowledge, apparently, is more tacit for some people than for others. Incentives, too, clearly matter: when circumstances place a great premium on effective articulation, remarkable things can sometimes be accomplished. For example, it has been established in occasional emergency situations that it is not impossible to convey by radioed verbal commands enough information on how to fly a small plane so that a person who lacks a pilot’s skills can bring the plane in for a landing.
As we observed previously, a variety of terms have been used in the social science literature to refer to concepts closely related to “skill.” It is interesting and somewhat curious that the array of terms employed in this connection includes several whose connotations are to a degree adverse to tacitness. The above passage from Miller, Ga lanter, and Pribam is indicative of the fact that their notion of a “Plan” is intimately related to the usual idea of a skill, and also to the idea that words may not suffice to communicate a plan. Yet the word itself, in ordinary usage, usually refers to something that is articu lable and capable of being represented symbolically. A similar obser vation holds for “program,” a term favored by March and Simon, among others.
Schank and Abelson employ the term ” script” to refer to “a struc ture that describes appropriate sequences of events in a particular context, . . . a predetermined, s tereotyped sequence of actions that defines a well-known situation” (Schank and Abelson, 1977, p. 41) . As with “plan,” the connotations of “script” cl early favor the notion that the knowledge involved can be articulated. Nevertheless, scripts turn out to rese mble skills rather closely, as the reference to “stereo typed sequence” suggests. To the extent that there is a distinction, the key to it lies in the fact that Schank and Abelson are concerned above all with the process by which natural language is understood. This concern entails a focus upon the successful use of language: the inquiry relates to how this is accomplished in human beings and how it might be accomplished by a computer. A vast realm of tacit knowledge is nevertheless implied by the computer programs that Schank and Abelson devise to represent the processes of under standing. They are well aware ot but do not focus upon, the fact that these programs imply a great deal of information processing that is not part of the conscious activity of a human being who is trying to understand. Indeed, were it not the case that the inferential processes they attempt to model are imperfectly accessible to conscious thought, the modeling task would be trivial and unworthy of the attention they bestow upon it. Thus, it seems that their approach to understanding of language does parallel Polanyi’s characterization of skill as involving “the observance of a set of rules which are not known as such to the person following them.”
In an important sense, the researcher who is attempting to build a computer model of human psychological processes is in a position analogous to that of a student attempting to learn a skill from an in structor. Both are betting that language can serve to communicate useful guidance to the underlying structure and details of a complex performance: the student seeks such guidance from his instructor and the researcher seeks it from his subject or, introspectively, from himself. Both would like to know how the thing is really done, the student for the sake of being able to do it and the researcher for the sake of being able to explain how it is done. Both are aware that, to the extent they experience difficulty in achieving their goals, lan guage is an imperfect tool for conveying the information they need. Language can communicate a framework, but a great deal of filling- in remains to be done after the resources of language are ex hausted; much of the filling-in involves labol’ious trial-and-error search. Perhaps both the student and the researcher tend to suffer from ambivalence regarding the limitations of language. Both hope that words will smooth their individual paths to achievement; both know that there is no distinction in the achievement if the path is too smooth.
For many reasons, it is important to try to identify the deter minants of the II degree of tacitness”-that is, the considerations that make tacit knowledge a more important part of the picture in some cases than in others. As a preliminary step in this direction, we will consider here the sources from which the limits on the articulation of knowledge derive. Such limits seem to arise in three distinguishable ways.
There is, first of all, a limit imposed by the feasible time rate of in formation transfer through symbolic communication, which may be well below the rate necessary or appropriate in the actual p erform ance. In the case of serving a tennis ball or performing a gymnastic stunt, the law of gravity imposes a tight constraint on the rate at which critical pOliions of the maneuver are performed. Thus, although step-by-step description is possible, and pretrial instruc tion and posttrial criticism are both helpful, it is not realistic to offer detailed instruction during an attempt. And although the learner can attempt to store pretrial instruction in memory and consciously re trieve it as the action is performed, the effectiveness of this tactic is severely limited by the speed and simultaneity of the information processing required. Ultimately, therefore, the learner has to work out the details of the coordination problem for himself. His knowl edge of those details remains tacit, is recollected without conscious awareness, and is probably no more susceptible to articulation than his instructor’s corresponding knowledge was.
Time-rate consid erations also figure, though in a somewhat dif ferent way, in learning touch typi ng or piano playing. In these cases, it is at least possible to enhance the role of articulation and of con scious awareness by slowing the time rate of the performance, and this fact is commonly exploited in learning. Nevertheless, the details of an accomplished performance are tacit: it is not the case that one can learn to perform the task on the “slow” setting and then simply push the /lfast” button to produce an expert performance.
A second consideration that limits the articulation of the knowl-edge underlying a skill is the limited causal depth of the knowledge. Polanyi’s swimming example illustrates the point that possession of a skill does not require theoretical understanding of the basis of the skill . In fact, it seems quite clear for all psychomotor skills that the actual mode of storage of the knowledge in the nervous system makes no use of the terms in which physicists, physiologists, and psychologists would describe the skilled performance . Yet this does not imply that an attempt to articulate the basis of the skill would not benefit from the availability of this terminology. Perhaps some no vice swimmers would be helped by Polanyi’s brief explanation of the body’s buoyancy. More generally, we may note that a skilled per formance takes place in a context defined by the values of a wide range of variables relevant to the performance; these may include as pects of the performer’ s physical state, as well as conditions of air pressure and lighting, gravitational forces, and so forth. The per former need not be aware of the existence of all of these variables, let alone of their relevance to the performance. This means that the per former simply relies upon these variables being in acceptable ranges , and is in no position to describe what it is that he relies upon. Should the values of some of the variables change so that the con straints are violated, the limited causal depth of the knowledge in volved will i mpede or prevent effective adjustment to the change.
The third aspect of the limitation of articulation is the coherence aspect-that of the whole versus the parts. Efforts to articulate “com-plete” knowledge of something by exhaustive attention to details and thorough discussion of preconditions succeed only in producing an incoherent message. This difficulty is probably rooted to a sub stantial extent in the related facts of the linear character of language-based communication, the serial character of the “‘central processor” of the human brain, and the relatively limited capacity of human short-term memory. Given these facts, the possibilities of articulating both the details and the coherent patterns they form the relationships among the details- are necessarily limited. At a given point in a text, a passage is encountered in a context estab lished by nearby passages; to convey the fact that it is also meaning fully connected to other parts of the text requires more words, and places demands on the reader’s memory. Similarly, it is difficult to form coherent three-dimensional mental images from exposures to a number of two- dimensional cross-sections of an object. To cope with these limitations of human powers of articulation and symbolic in formation processing, a variety of aids are employed that present in formation about patterns and structures directly to the eyes- aids such as photographs, diagrams, graphs, flowcharts, and holograms. There is a rapidly advancing technology of such aids.
In short, much operational knowledge remains tacit because it cannot be articulated fast enough, because it is impossible to articu late all that is necessary to a successful performance, and because lan guage cannot simultaneously serve to describe relationships and characterize the things related. This observation provides us with at least a starting point for assessing the relative significance of tacit knowledge in different situations. The knowledge contained in the how-to-do-it book and its various supplements and analogues tends to be more adequate when the pace of the required performance is slow and pace variations are tolerable, where a standardized, con trolled context for the performance is somehow assured, and where the performance as a whole is truly reducible to a set of simple parts that relate to one another only in very simple ways. To the extent that these conditions do not hold, the role of tacit knowledge in the per formance may be expected to be large.
Finally, it should be emphasized that costs matter. Whether a par ticular bit of knowledge is in principle articulable or necessarily tacit is not the relevant question in most behavioral situations. Rather, the question is whether the costs associated with the obstacles to articulation are suffici ently high so that the knowledge in fact re mains tacit.
Source: Nelson Richard R., Winter Sidney G. (1985), An Evolutionary Theory of Economic Change, Belknap Press: An Imprint of Harvard University Press.