Chapter 1 — From the Workshop to the Laboratory
Prologue: What an Economy Can Do with What It Knows
Imagine a public technical institute with enough money to buy a new instrument, but no provision for the person who will keep its measurements comparable from one year to the next. The purchase has an identifiable supplier, a delivery date and a photograph for the annual report. Continuing technical competence is harder to present. It includes learning which results to distrust, maintaining reference materials, training a replacement for someone who leaves and answering questions from people whose work does not fit the institute's original specification.
This is a hypothetical budget decision, not a reported failure at a particular institute. Its purpose is to make a difficulty visible before we attach it to a celebrated invention. An instrument can arrive while the capacity to use it remains incomplete. Equally, withholding the purchase until every future staffing question has been settled could leave an excellent group working with equipment that prevents it from answering an important question. Neither the photograph nor the staffing plan tells us enough. We need to understand the work each expenditure makes possible and the other commitments on which it depends.
The people making this decision may agree that innovation matters and still disagree about almost everything that follows. One wants room for inquiry whose application cannot yet be specified. Another needs results that a manufacturer can reproduce. A third wants to know whether smaller organizations will ever get access. Their disagreement concerns purposes, timescales and authority as well as money. A larger research budget would relax one constraint without settling those questions.
Innovation economies are built out of such incomplete agreements. People have to carry uncertain work through organizations that answer to different constituencies. A laboratory may value an unexpected finding, a production manager may need stable tolerances, and a public buyer may be unable to purchase either until someone can demonstrate how the proposed improvement will perform in service. These are intelligible responsibilities. When they do not connect, even capable people can leave one another with work that cannot be used.
There is also a danger in asking everything to lead immediately to a purchase. Inquiry can change the questions worth asking. A reliable account of why a proposed technique fails may be more useful than a pleasing demonstration that conceals the circumstances of its success. An economy needs people who can pursue those findings without having to describe each one as the beginning of a market. It also needs ways for findings to escape the group that first understood them. Protecting inquiry and making knowledge usable are demanding tasks in their own right; declaring a commitment to both does not arrange their relationship.
Our vocabulary can obscure the distance involved. In the fourth edition of the *Oslo Manual*, the OECD and Eurostat distinguish innovation activities from implemented innovations. A business innovation must differ significantly from the firm's previous products or processes, but it need not be new to the world, commercially successful or beneficial to society. Those definitions help make survey answers comparable; they are not a verdict on what a society should want. A count of innovations therefore cannot, on its own, establish an improvement in collective well-being.[0-1]
The question pursued here is more specific: how do people acquire the ability to turn knowledge into capabilities that others can use, inspect and improve? A capability is more than a document describing a result. Someone must be able to perform the relevant task with available materials, skills and equipment, recognize when performance is deteriorating, and make a judgment about what to change. Depending on the task, that ability may belong to a single craftsperson, a team, a firm or a much wider network. Calling knowledge public does not tell us who possesses it in this operational sense.
Return to the hypothetical institute. Its staff could publish every measurement method and still have to decide how an unfamiliar user will learn to prepare a sample. If the preparation changes the result, a downloadable procedure may leave out precisely the judgment that matters. Opening a training place takes time from other work. Charging for access can pay for that time while excluding a group with a valuable problem and little cash. Offering free access can move the shortage into a queue. There is no universally correct fee concealed in the word openness. The consequences depend on who needs access, what they must learn and which costs somebody is prepared to carry.
Some problems can be resolved by a better description or a more convenient transaction. Others require a new relationship. A manufacturer who repeatedly sends technicians to the institute may learn to identify a difficulty before sending a sample; the institute may learn that its convenient laboratory preparation is unrepresentative of production. That mutual adjustment could improve both tasks. It could also make the institute excessively attentive to its largest customer. Access arrangements shape what people discover as well as who receives an already completed discovery.
That is the sense in which this book uses the word structural. We will follow the arrangements through which people can do work together: who can enter, who pays for an unfinished task, who can challenge a result, and who has enough authority to change a procedure. Structures do not supply a hidden explanation for every outcome. They alter the opportunities for judgment and the consequences of exercising it. A well-funded institution can make a poor choice. An outsider can discover something that its established experts missed. The interesting question is what happens next, especially when accepting the outsider's result would disrupt commitments on which respected people have built their careers.
History is essential to that inquiry because present institutions were made under particular circumstances. A workshop did not become obsolete when someone organized a laboratory. Nor does putting researchers on a company's payroll settle the relationship between investigation and commercial application. Reading the original records allows us to ask what an arrangement was meant to achieve without treating its authors' ambitions as evidence that it succeeded. Later chapters extend the inquiry across firms, cities and international projects, where no participant commands all the complementary work.
AI and widely shared software bring new possibilities into this picture. Their importance will depend partly on the tasks they improve and partly on what users must still supply. Generating more candidates may increase the demand for selection and testing. A reusable tool can widen participation while increasing dependence on maintainers whose work users rarely see. The later chapters follow those decisions: which possibilities merit investigation, and what continuing work makes a shared tool dependable.
At the hypothetical institute, the engineer asking for new equipment may be right. So may the technician who asks for time to train a successor. Their claims cannot be ranked by labeling one innovative and the other routine. To judge them, we have to see what each would enable, what it would displace and who could carry it through. That means following the work beyond the expenditure that is easiest to count.
Notes and Sources
[0-1] OECD/Eurostat, *Oslo Manual 2018*, fourth edition, chapter 3, especially paragraphs 3.4–3.18. These are institutional measurement definitions, not causal estimates or a welfare ranking. https://www.oecd.org/en/publications/oslo-manual-2018_9789264304604-en/full-report/component-9.html
Chapter 1 — From the Workshop to the Laboratory
Robert Hooke's *Micrographia*, published in 1665, contains a procedure for making lead shot among its observations of the minute world. Hooke acknowledged permission conveyed through Robert Moray to print it. The account specified equipment and directed operators to inspect the falling shot and adjust the heat. It offered an explanation of the round shape while leaving that explanation open to further inquiry. Elsewhere in the preface, Hooke acknowledged the makers and grinders of his glasses. Practical production, instruments, publication, and uncertain explanation occupied the same work.[1-1]
A capable maker may recognize a fault, correct it, and produce an excellent object. Someone who buys that object acquires its service. Someone who watches the maker may acquire part of the skill. Neither transaction necessarily gives a distant investigator the means to explain the correction, test its limits, or reproduce it with different materials. Knowledge can be highly effective within a particular relationship and remain difficult to carry beyond it.
There is no need to force this record into a story of one kind of knowledge defeating another. A procedure can travel while its explanation remains unsettled. An explanation can improve while the most dependable way to carry out the procedure still depends on an experienced hand. The interesting question is where the dependence lies and whether another person has a reasonable route to acquiring it. Calling an activity scientific does not answer that question. Nor does calling it craft establish that it cannot be examined or communicated.
What a procedure has to carry
A usable instruction has at least two audiences. One wants to achieve the result under conditions reasonably close to the original conditions. The other wants to discover how far those conditions can change. The first needs a reliable sequence. The second also needs to know which features of the sequence matter. These audiences overlap, but a document written entirely for one may serve the other poorly. A production recipe can include a precaution whose reason nobody has established; an experimental paper can establish a relationship without providing a complete production method.
Some instructions specify an action; others ask the user to make a judgment. A quantity, distance or duration can be stated directly. Recognizing an acceptable surface or the right moment to stop depends on perception. A learner can acquire that judgment by comparing results with an experienced practitioner, but a distant reader lacks the comparison. More description may help, or the missing resource may be a reference specimen, supervised practice or both.
The cost of transfer therefore depends partly on what a recipient already knows. A description that seems complete to its author may omit steps shared by everyone in the author's immediate circle. When that circle expands, the omissions become visible. This does not mean the original author concealed anything. Communicating every movement and assumption would make many instructions unusably long. The practical task is to identify the assumptions most likely to fail at the receiving end and decide how to make them inspectable.
One response is to send an object with the description. A drawing can identify geometry; a specimen can establish a comparison; a measuring instrument can make an otherwise vague threshold more definite. Each response creates another dependency. The specimen may deteriorate. The drawing may omit the finish that governs performance. The instrument may drift or respond differently in another environment. Transfer is a sequence of judgments about which dependencies are tolerable, rather than a single act of putting an idea on paper.
This sequence also changes who can participate. A recipient who can read the description but cannot obtain the required material has acquired information without the means of execution. Another may possess the equipment but lack time to resolve its irregular behavior. The same publication can be a practical resource for one reader and little more than an announcement for another. Counting copies, citations, or visitors would not distinguish those experiences. An account of innovation has to follow the receiving work as well as the original act of disclosure.
Making a troublesome effect into a question
In his 1798 report on heat generated by friction, Benjamin Count of Rumford began from cannon boring at the military arsenal in Munich. He compared the heat capacities of chips and thin pieces of the same gun metal, then described a modified boring arrangement for investigating the heat produced. His accounting explicitly omitted heat accumulated in the wooden enclosure and heat dispersed during the experiment. The report presents a production setting turned into an inquiry with comparisons and acknowledged limits.[1-2]
That movement suggests a second distinction: noticing an effect and creating a testable disagreement about it. An effect may be familiar without being well understood. To investigate it, someone must identify competing explanations that would lead to different observations. The design of the comparison matters because it determines what a result can settle. A striking demonstration can establish that something happens while leaving several explanations available. A quieter comparison may eliminate one explanation without looking especially impressive to an audience.
The transformation also requires permission to interrupt ordinary work. Production equipment exists to deliver an output. Investigating it can consume material, alter its condition, occupy operators, or delay an order. Even when no expensive new instrument is required, the inquiry has a cost. Someone controls access to the equipment and someone bears the interruption. A useful observation can remain unexamined because the people who recognize it cannot authorize the necessary pause. This is a general organizational possibility, not a claim about an undocumented refusal in the Munich arsenal.
Separating these decisions helps explain why access to a workshop is not equivalent to access to an experiment. The workshop supplies material and operations. Investigation adds a question, a comparison, and a way of interpreting an outcome. Those additions may be modest or demanding. They can also require authority different from the authority to operate the equipment. A skilled operator might be allowed to adjust a process to meet an order while being forbidden to vary it in a way that risks a batch. The knowledge needed to make the adjustment and the permission needed to investigate it may belong to different people.
An innovation economy benefits when those people can reach one another at a reasonable cost. That does not require every workshop to become a research institution. It requires some route through which an unresolved technical question can obtain attention, equipment, and a competent comparison. The route might involve a practitioner investigating alone, a visiting investigator, a shared facility, or a specialized group. Its quality depends on whether the question survives translation. If the recipient receives only a tidy account of success, the irregularity that made the problem worth studying may disappear.
The opposite problem is equally real. Investigation can absorb effort without producing a useful distinction. Collecting more observations is not automatically progress if no one can say what different results would imply. A comparison needs enough discipline to expose a claim to possible failure. Otherwise the apparatus may produce an elaborate record that cannot guide a subsequent decision. Laboratories provide opportunities for such discipline; their existence does not guarantee its exercise.
Precision is organized work
James Prescott Joule's 1850 paper on the mechanical equivalent of heat shows another part of the process. He described thermometers supplied by named makers, comparisons between their scales, and reference conditions. He paired friction experiments with trials used to assess heat exchange with the surroundings. In discussing solid friction, he also reported difficulty controlling the regulating lever and accounted for vibration and sound. The paper's numerical ambition depended on instruments, practiced operation, and corrections that could themselves be examined.[1-3]
A measurement is useful partly because it tells someone what to compare. Two investigators may both report a temperature change and still disagree because their instruments, reference points, or corrections differ. Without a record of those choices, repeating the experiment may reproduce the uncertainty rather than resolve it. More decimal places do not repair an unidentified difference in procedure. A precise-looking result can be difficult to challenge for exactly the reason it is difficult to use: its dependencies remain hidden.
Here the laboratory offers a particular service. It can make time and equipment available for examining the measurement process itself. An operator producing goods ordinarily has reasons to stop once the output meets the required standard. An investigator estimating a relationship may need to continue after that point, changing instruments, examining disturbances, or repeating a trial that has already produced a satisfactory visible effect. The additional work has no necessary counterpart in a customer's immediate demand. It needs a purpose and a source of support of its own.
The result of that work is partly a boundary around the claim. A finding that holds under specified conditions can be more useful than a sweeping claim whose limits are unknown. The boundary allows another person to decide whether to apply the result directly, commission a further test, or set it aside. It also makes disagreement more informative. A failed attempt outside the stated conditions poses a different question from a failed attempt within them. An economy that preserves those distinctions can direct subsequent effort more intelligently.
This does not make every correction equally worthwhile. Measuring an effect more accurately consumes resources, and the additional accuracy may have little bearing on the intended use. The appropriate standard depends on the decision. A rough estimate may be sufficient to reject an impractical design; a much tighter estimate may be needed before committing to a costly system. The case for greater precision should therefore identify the uncertainty it will reduce and the action that uncertainty currently prevents. Precision is a capability with a cost, not a virtue that can be pursued without limit.
Nor should precision become a test of who deserves to speak. A practitioner who cannot yet measure an irregularity may still have identified a consequential problem. If a research group treats only already standardized observations as admissible, it may exclude exactly the questions for which its equipment would be valuable. The useful relationship allows an observation to enter provisionally, followed by work that establishes what can actually be claimed. Demanding final evidence at the point of entry would confuse the reason for an inquiry with its eventual result.
Consider an elementary measurement example, constructed here to distinguish the work involved. Two thermometers can disagree by a constant amount and still report the same change between a beginning and an end. If one reads two degrees high at both points, that offset cancels in the subtraction. A scale that expands the interval between readings creates a different problem: the reported change itself is wrong. Checking a single reference point cannot distinguish the two. A recipient needs to know what was compared across the range actually used.
The example matters for transfer because agreement on a final number can have several causes. Investigators might use well-calibrated instruments. They might share the same error. Or different errors might happen to offset one another in the reported result. Agreement is informative, but the route to agreement determines how much reliance it deserves. A method that makes its comparisons explicit lets the next investigator decide which possibility needs examination. Merely repeating the final number makes that decision harder.
There is also a difference between detecting a change and accounting for its source. In a constructed trial, a vessel becomes warmer while an operation is performed. A separate period with the operation stopped might reveal that the surroundings also warm the vessel. Subtracting that estimate can improve the inference only if the stopped period is a relevant comparison. If the surroundings change sharply between periods, the correction may introduce another uncertainty. The calculation is simple; designing a comparison that makes the subtraction meaningful is the demanding part.
Such examples identify why a portable finding contains more than a formula. The recipient needs to recognize which quantities belong in the formula and how they were obtained. Training someone to do the arithmetic is different from training that person to judge whether a comparison applies. Both can be taught, but the second requires encountering cases in which superficially similar observations have different implications. A laboratory capable of producing a result once has not necessarily developed a method capable of teaching that judgment.
The original investigator also needs a stopping rule for calibration work. If uncertainty in the instrument is much smaller than variation in the material, another refinement of the scale may leave the important question unchanged. Effort could be better spent characterizing the material or altering the procedure. This judgment links measurement to the intended decision. It prevents precision from becoming an isolated competition and gives the next participant an explanation of why some uncertainties were reduced while others remain.
Units create a related obligation. Translating a number from one scale to another is straightforward only when the scales' relationship is established over the relevant range. Repeating a conversion rule can conceal an uncertainty in the original measurement or reference. A recipient should be able to distinguish a new observation from a restatement of an old one in different units. Otherwise two apparently independent confirmations may be the same evidence counted twice. The transferable contribution includes the provenance of the number, not merely the convenience of its presentation. This is especially consequential when results pass through several people before reaching the person who must act on them.
The residual skill
The historical records considered here do not support the idea that transferring knowledge removes the need for skill. They direct attention to a more practical question: which skills must the next participant possess, and how can that person acquire them? A written method may reduce the amount of apprenticeship required while increasing the importance of instrument handling. A standard specimen may simplify comparison while creating a need to preserve and verify it. Improving one part of a method can move the difficult judgment elsewhere.
There is a temptation to call whatever remains unwritten mysterious. That is rarely a useful stopping point. Some judgments remain unwritten because no one has taken time to describe them. Others are difficult to describe because the relevant variation is continuous and the signs overlap. Still others involve an active response to changing conditions. Each difficulty suggests different work. Better records may help the first; reference examples and coached practice may help the second; a more tolerant process design may help the third. Merely announcing that knowledge is tacit does not choose among them.
This analysis also limits what can be inferred from a successful demonstration. The demonstrator may have compensated for small variations without mentioning them. A second person may follow every explicit instruction and fail because those compensations were never transferred. Treating the failure as incompetence would close inquiry prematurely. Treating it as proof that the original result was false would be equally premature. A useful response asks which relevant actions, materials, and judgments differed and whether the claimed result depended on them.
A method becomes more robust when it tolerates some differences without losing its function. That can be economically important even if it adds nothing to the explanation of the underlying phenomenon. Replacing a sensitive operation with a more forgiving one may allow less experienced users to succeed, reduce the frequency of supervision, and widen the range of acceptable inputs. These are changes in capability. They should not be dismissed because they lack the drama of a new principle, and they should not be credited to a scientific finding before the development work has been done.
There are also cases where preserving a skilled practice is the sensible choice. If the work is infrequent, highly variable, or difficult to evaluate until completed, the cost of fully specifying it may exceed the benefit. A small group may reasonably rely on training and direct observation. The limitation becomes consequential when outsiders assume that a short description has made the practice universally available. An honest account can state what has been shared and what still requires contact with experienced people. That statement improves the recipient's planning even when it leaves a substantial learning cost.
Building a place for unresolved work
Organized investigation changes the continuity of attention. A problem that survives one failed trial needs somewhere to remain. Someone must preserve the apparatus, identify the material used, retain the observations, and explain to the next participant why an apparently obvious approach was abandoned. Without that continuity, a succession of capable people may repeatedly return to the same starting point. A dedicated room helps only if people can keep using it and can recover what earlier work established.
Consider a hypothetical handover of an unfamiliar experimental setup. Equipment may need repair, supplies may vary, and a notebook may preserve dimensions without explaining the order of adjustment. The next participant may need to operate the apparatus while the previous participant is still available. That overlap costs time, but it can reveal omissions before the only person able to explain them leaves.
The same reasoning applies to records. Recording everything is expensive and can bury the few distinctions a successor needs. Recording only success erases the path by which alternatives were rejected. A useful record preserves enough of the decision structure to make the next inquiry economical: what was tried, what happened, which uncertainty remains, and which features of the method are still suspect. Its value appears when someone uses it to avoid repeating an uninformative test or to design a more discriminating one.
This creates a reason for institutions to support activities whose return arrives elsewhere. The person who writes a careful account may never meet the recipient who benefits. The maker who improves an instrument may enable observations on a different subject. The group that identifies a measurement error may save another group months of effort while receiving little visible credit for a discovery. These are possible relations among contributions, not a claim that every act of disclosure creates a net social gain. Some methods prove unusable, and some carefully preserved questions never justify further work.
The economic issue is whether the arrangements permit valuable connections to be attempted without requiring each contributor to capture the entire eventual return. Here it arises in a relatively concrete form: can someone with a troublesome effect obtain the means to test it, and can someone elsewhere obtain the means to judge the result? A society can fail at either end. It can leave inquiry unsupported or leave its results inaccessible in practice.
The workshop and the laboratory consequently name overlapping sites of work rather than successive stages through which every economy must pass. Production can supply questions, material, and exacting practical tests. Organized inquiry can separate effects, establish comparisons, and make claims easier to inspect. Instrument makers and trained operators connect the two. The boundaries shift with the task; drawing them in advance can conceal the labor that makes a result portable.
A finding becomes a broader capability when another group can take responsibility for using it with some defensible understanding of its limits. The practical test is whether that group can distinguish a result it should reproduce from a condition it must first investigate. A recipe, an instrument, and an explanation can each help. None removes the need to find out which uncertainty now belongs to the person receiving the work.
Notes and Sources
[1-1] Robert Hooke, *Micrographia* (1665), preface and Observation VIII, lead-shot procedure. Gutenberg ebook 15491 transcription. The text supplies a procedure and qualified explanation, not evidence of its general uptake. https://www.gutenberg.org/cache/epub/15491/pg15491-images.html
[1-2] Benjamin Count of Rumford, “An Inquiry concerning the Source of the Heat which is excited by Friction,” *Philosophical Transactions* 88 (1798), 80–102, 81–83 and 93. https://academicweb.nd.edu/~powers/ame.20231/thompson1798.pdf
[1-3] James Prescott Joule, “On the Mechanical Equivalent of Heat,” *Philosophical Transactions* 140 (1850), 61–82, particularly 63–66 and 78–79. https://www2.lawrence.edu/fast/STONEKIM/COURSES/Joule_1850.pdf