Everyone would agree that in order to assess the impact of technical change on society in general or on any particular aspect of human activity, it is necessary to have some basis for forecasting. If new technologies fall upon us like a hailstorm or surprise us like an earthquake, there is little we can do as a society to master them or guide them for the common good. What we will argue here is that, in spite of the undeniable diversity of technologies, of the unpredictable nature of inventions and of the uncertain and risky nature of commercial innovations, there is a recognizable logic behind the main trends in technical change.

To develop the analysis we need a set of appropriate concepts for classification. The most basic is the Schumpeterian [1] distinction between invention, innovation and diffusion.

The invention of a new product or process occurs within what could be called the techno-scientific sphere and it can remain there forever. By contrast, an innovation is an economic fact. The first commercial introduction of an innovation transfers it into the techno-economic sphere as an isolated event, the future of which will be decided in the market. In case of failure, it can disappear for a long time or forever. In case of success it can either still remain an isolated fact or become economically significant, depending upon the degree of appropriability, its impact on competitors or on other areas of economic activity. Yet, the fact with the most far-reaching social consequences is the process of massive adoption. Vast diffusion is what really transforms what was once an invention into a socio-economic phenomenon.

So, inventions can occur at any time, with different importance and at varying rhythms. Not all of them become innovations and not all innovations diffuse widely. In fact, the world of the technically feasible is always much greater than that of the economically profitable and this, in turn, is much greater than that of the socially acceptable.

Thus, our focus must be innovation diffusion. Let us then establish a manner of classifying innovations which will help us understand the economic and social conditions for diffusion and will give us some insight into how meaningful trends in technical change can be discerned.
        

Incremental innovations are successive improvements upon existing products and processes. From an economic point of view, this type of change lies behind the general rate of growth of productivity, visible in the aggregate. The frequent increases in technical efficiency, productivity and precision in processes, the regular changes in products to achieve better quality, reduce costs or widen their range of uses, are characteristic features of the evolutionary dynamics of every particular technology. The logic guiding this evolution, called "natural trajectory" by Nelson and Winter [2] and "technological paradigm" by Dosi [3] , is analyzable and makes the course of incremental change relatively predictable. Given a technological base and the fundamental economic principles, it is possible to forecast with a reasonable degree of certainty that microprocessors, for example, will become smaller, more powerful, faster in operation, etc. Once catalytic refining was introduced, and knowing the profile of demand for oil derivatives, it was natural to expect that technological evolution would lead to successive improvements geared to increasing the yield of gasoline to the detriment of the heavier products with lower demand and lower prices. In the process industries, after the discovery of Chilton's Law, according to which doubling plant capacity only increased investment cost by two-thirds, it was easy to expect a trend towards obtaining those scale economies in a whole range of industries. So, the great majority of innovations occur in a continuous flow of incremental changes along expected directions.

A radical innovation, by contrast, is the introduction of a truly new product or process. As both Freeman[4] and Mensch[5] observe, due to the self-contained nature of the trajectories of incremental change, it is practically impossible for a radical innovation to result from efforts to improve an existing technology. Nylon could not result from successive improvements to rayon plants, nor could nuclear energy be developed through a series of innovations in fossil fuel electric plants. A radical innovation is, by definition, a departure, capable of initiating a new technological course. Although radical innovations are more willingly adopted when the previous established trajectory approaches exhaustion, they can be introduced at any point in time cutting short the life cycle of the products or processes they substitute. There are some radical innovations that give birth to a whole new industry. Television, for instance, not only introduced a manufacturing industry but also programming and broadcasting services, which in turn widened the scope of the advertising industry. In this sense, important radical innovations are at the core of the forces behind growth and structural change in the economy.
           

The combination of these two concepts allows us to visualize the evolution of a technology from introduction to maturity, as shown in Figure 1-A. Every radically new product, when it is first introduced, is relatively primitive. In the initial period there is much experimenting in the product and in its process of production, in the market and among the initial users. Gradually, it consolidates a position in the market and the main trends of its trajectory are identified. From then on, there is a sort of take-off for a period of successive incremental improvements in quality, efficiency, cost-effectiveness and other variables, which eventually confronts limits. At that point, the technology reaches maturity. It has lost its dynamism and its profitability. Depending on the type of product, this cycle can last months, years or decades; it can involve a single firm, dozens of firms or thousands. As the technology approaches maturity, there is often a shakeout, leaving only a few producers. There is also a high likelihood that, at maturity, the product will be replaced by another or the technology will be sold to weaker producers with lower factor costs (such as happened in deployment of mature industries to the Third World in the late 1960's and 1970's).

Thus forecasting in relation to single technologies is on relatively firm ground, and is, in fact, quite common in the daily practice of engineers, managers and investors. For each individual product or process, incremental change is not random and its destiny, unless another radical innovation appears, is to reach maturity and exhaustion. There are, then, moments of discontinuity and periods of continuity in the evolution of each individual technology.

This, of course, does not lead to long waves. Individual innovations -radical and incremental- are constantly happening in products and processes, in different industries and different places, some are minor some are major, some have a long life others a short one. Indeed, if technologies developed isolated from each other, the rise of the life cycle of some technologies would counter the maturity and decline of others. But technologies grow in systems.
           

Freeman[6] has defined technological systems as constellations of innovations, technically and economically interrelated and affecting several branches of production. Rosenberg[7] has described the way in which some innovations induce the appearance of others. Breakthroughs that increase the speed of operation of machine-tools, for instance, induce innovative efforts in cutting alloys capable of withstanding greater temperatures and speeds and, in general, incremental trajectories in a product, process or branch of industry tend to encounter bottlenecks which become incentives for innovations -even radical ones- in other industries. Nelson and Winter[8] identify generic technologies, whose natural trajectory of evolution encompasses that of a whole set of interconnected radical innovations.

In petrochemical technology, for instance, one can identify several distinct but related systems: synthetic fibers, which transform the textile and garment industries; plastics, whose multiple impact, in the form of structural materials, generates whole new lines of equipment for extrusion, molding and cutting, and whose versatility transforms the packaging industry and opens a vast universe of innovations in disposable products; and so on.

From the vantage point of a technological system, then, there is a logic, which joins successive interrelated radical innovations in a common natural trajectory. Once this logic is established for the system, it is possible to forecast a growing succession of new products and processes each of which, taken individually, appears as a radical innovation, but when located within the system can be considered as an incremental change. The series of durable consumer goods, made of metal or plastic with an electric motor, which begins with the vacuum cleaner and washing machine, goes through food processors and freezers, to later approach exhaustion with the electric can-opener and the electric carving knife, is a banal example of this type of logic in the area of products. The succession of plastic materials with the most diverse characteristics, based on the same principles of organic chemistry, is an example in the field of intermediate products with enormous impact in generating innovations in the user industries. The "Green Revolution", with the introduction of growing families of oil driven agricultural machinery, together with multiple petrochemical innovations in fertilizers, herbicides and pesticides, is an example of the coherent evolution in the logic of a productive system.

The widespread impact of a new technological system stems from the "wide adaptability"[9] of the contributing innovations and from their multiple character. They are not merely technological. Each technology system brings together technical innovations in inputs, products and processes with organizational and managerial innovations. Further still, they can induce important social, institutional and even political changes. The technological constellation of the "Green Revolution" led to single-crop farming in great expanses of land and induced changes in the organization of production and distribution as well as in the structure of ownership. The automobile, the assembly line, the networks of parts suppliers, distributors and service stations, suburban living and commercial centers, are only some of the elements of the technical, economic and social constellation gradually built around the internal combustion engine.

Yet technology systems, in a manner similar to individual technologies, eventually exhaust their potential for further growth and improvement. For a long time, a technological system provides multiple and growing opportunities for innovation and investment in complementary products, services or supplies. But, the time comes when the system loses technological and market dynamism, reaches maturity, threatens the growth and profits of most of the firms involved and, therefore, stimulates a search for radical new products that will serve as the core of other new technology systems.

So, at the level of technology systems, we re-encounter the same phenomena of continuity and discontinuity in evolution. Again, at first sight, there is no reason to expect long waves to occur because of limits in the life cycle of technology systems. As with individual innovations, one could imagine a constant process of counterbalancing of the growth and decline of different systems in different parts of the economy. This would be the case if systems developed in isolation, but technology systems grow in interconnection with each other and with the surrounding economic, cultural and institutional environment.
     

In the early 1970's, it was widely agreed (and feared) that the automobile industry had reached maturity. Its markets had lost dynamism and grew extremely slowly -if at all-, inventories piled up, productivity stagnated and profits were threatened. Many experts declared that automobiles had become "commodities" and the future was seen as complete standardization by moving towards the "world car:" Engines would be produced in one country, gear boxes in another, bodies in the next, and so on, in order to increase productivity through maximizing economies of scale. This was the way imagined by the mentality of the time to confront the maturity of that technology system.

Few could foresee what happened. Japanese industry developed a distinctly different way of organizing production and markets, which, at first, threatened to overtake much of the world automobile industry but, instead, led to a thorough revamping of all firms and their forms of insertion, competition and interrelation. In the end, through a synergistic combination of the new managerial style and the introduction of information technology in the processes and in the products, in administration and in markets, the industry was completely renewed and set on a different and very dynamic trajectory of incremental innovation [11] .

So, maturity does not inevitably end in the marginalization of a system, nor is it necessary that a radical innovation in the core product itself should come to the rescue and replace the previous mature product. Both can occur and sometimes do occur. What is more likely to take place, especially at those times -such as the 1970's- when many inter-related systems tend to come to maturity more or less simultaneously, is that a general solution appears in the form of a technological revolution. What happens then, is the diffusion of a new set of generic technologies, capable of rejuvenating and transforming practically all existing industries, together with the creation of a group of new dynamic industries, at the core of radically new technology systems. These are the technological revolutions described by Schumpeter [12] as "creative gales of destruction." They have occurred about every fifty or sixty years and it is this phenomenon that lies at the root of the so-called long waves in economic growth.

Schumpeter and many others after him [13] have emphasized the powerfully dynamic nature of each of those great waves of new technologies as well as their capacity to profoundly modify the world around them. Society has recognized their overarching influence by referring to the periods when these great technological changes have diffused as the Industrial Revolution, the Railway Era, the Age of Electricity and the Age of the Automobile. The industries at the core of these revolutions do, indeed, become the propellers of growth for a considerable length of time. They also lead to the proliferation of whole new industries and services complementary to the production and use of the new products, as was discussed above for technology systems of major importance.

Yet, what we are suggesting is that they do much more than that. Technological revolutions change the "common sense" criteria for engineering and business behavior across the board. In fact, in our view, each technological revolution merits that name, not only for the importance of the new industries it ushers in and the new technical possibilities it opens but also -and perhaps mainly- because it radically modifies the "best practice frontier" for all sectors of the economy.

Each of these revolutions is, in fact, a constellation of technological systems with a common dynamics and including a set of generic technologies of widespread applicability. Its diffusion across the length and breadth of the productive sphere tends to encompass almost the whole of the economy and ends up transforming the ways of producing, the ways of living and the economic geography of the whole world.

Such all-pervasive revolutions generate, therefore, massive and fundamental changes in the behavior of economic agents. Yet, what type of mechanism would be capable of serving as guiding force for a shift of this sort?