driven mainly by for-profit drivers but there has been also an as yet incomplete convergence toward AKST relationships, arrangements and processes that foster innovations supportive of socially inclusive and ecologically sustainable and productive agricultures.

2.1.4.1 Changes in perspective: from technologies to innovations

The proposition that technical change could be a major engine of economic growth was demonstrated in the 1950s (Solow, 1957). Later analysis of empirical evidence showed that small-scale producers, although handicapped by severe constraints, made rational adaptations over time in their practices and technologies in response to those constraints. Insofar as externally introduced technology released some of the constraints, technology could become a driver of significant change (Schultz, 1964). The Green Revolution subsequently appeared to vindicate the analysis and it quickly became dominant in the agricultural economics profession (Mosher, 1966).The model that this analysis pointed toward is the dominant way of organizing knowledge and diffusion processes, i.e., "the transfer of technology model" (2.1.2) (e.g., Chambers and Jiggins, 1986). It is known also as a policy model, variously as "the agricultural treadmill" (e.g., Cochrane, 1958) and "the linear model" (e.g., Kline and Rosenberg, 1986); and its role in policy is assessed here. In its simplest form it recommends technology supply-push, i.e., developing productivity enhancing component technologies through research for delivery, transfer, or release to farmers, the "ultimate users".

     The model emerged in a specific historical context, the American Midwest in the decennia after WWII (Van den Ban, 1963); similar models were elaborated from empirical findings in other economic sectors. Although these mechanisms driving the model's impact are familiar to economists, they are not necessarily as familiar to others so the persistence of technology supply push as the dominant policy model for stimulating technology change in agriculture warrants a full explanation of the mechanisms. In the case of agriculture the empirical data robustly confirm the following features:

1. Diffusion of innovations. Some technologies diffuse quite rapidly in the farming community after their initial release, typically following the S-curve pattern of a slow start, rapid expansion and tapering off when all farmers for whom the innovation is relevant or feasible have adopted. The classic case is hybrid maize in Iowa (Ryan and Gross, 1943). Diffusion multiplies the impact of agricultural research and extension effort "for free". But diffusion is mainly observed ex-post: it is difficult to predict (or ensure) that it will take place (Rogers, 2003).

2. Agricultural treadmill. The treadmill refers to the same phenomenon but it focuses on the economics (Cochrane, 1958). Farmers who adopt early use of a technology that is more productive or less costly than the prevailing state-of-the-art technology, i.e., when prices have not as yet decreased as a result of increased efficiency, capture a windfall profit. When others begin to use the new technology, total production increases and prices start to fall. Farmers who have not yet adopted

the technology or practice experience a price squeeze: their incomes decrease even if they work as hard as before. Thus they must change; the treadmill refers to the fact that the market propels diffusion: it provides incentives for early adoption and disincentives for being late.

3. Terms of trade. A key underlying aspect of the treadmill is that farmers cannot retain the rewards of technical innovation. Because none of the thousands of small firms who produce a commodity can control the price, all try to produce as much as possible against the going price. Given the low elasticity of demand of agricultural products, prices are under constant downward pressure. During the last decennia, the price of food has continuously declined both in real and relative terms (World Bank, 2008). The farm subsidies in the US and Europe can be seen as a necessary cost for societal benefit without rural impoverishment.

4. Scale enlargement. In the tail of the diffusion process, farmers who are too poor, too small, too old, too stupid or too ill to adopt eventually drop out. Their resources are taken over by those who remain and who usually capture the windfall profits. This shakeout leads to economies of scale in the sector as a whole.

5. Internal rate of return. Investing in agricultural research and extension to feed the treadmill has a high internal rate of return (Evenson et al., 1979). The macro effects of relatively minor expenditures on technology development and delivery are major in terms of (1) reallocating labor from agriculture to other pursuits as agriculture becomes more efficient, (2) improving the competitive position of a country's agricultural exports on the world market, and (3) reducing the cost of food. An advantage is that farmers do not complain. Their representatives in the farmers' unions are among those who capture windfall profits and benefit from the process, even though in the end the process leads to loss of farmers' political power as their numbers dwindle to a few percent of the population. The treadmill encourages farmers to externalize social and environmental costs, which tend to be difficult to calculate and hence usually are not taken into account. One may note here that this process, first described at the national level in the case of the USA, also explains the growing gap in the productivity of agricultural labor between industrialized and developing countries and that it leads to overall efficiencies in production and reduced prices for consumers, outcomes that have favored its persistence as a dominant policy model.

However, other business analysts and social scientists throughout the period under review have stressed the concept of innovation rather than mere technical change as a measure of development. The evidence that technical change itself requires numerous, often subtle, but decisive steps before an adoption decision is made reinforced this view (Rogers, 1983). Others pointed to biophysical, sociocultural, institutional and organizational factors such that when the same technology is brought into use in different contexts the effects vary (Dixon et al., 2001). Recently more emphasis has been given to development of "best fit" technology