existing patterns of inequality (Gao and Li, 2006).The history of broadcast radio suggests that over time the "digital divide" may become narrower. Issues of the quality and relevance of the information available are likely to become more important than those of access and ability to use the technology.

2.1.3 Science processes

Science processes are those involved in the creation and dissemination of scientific knowledge; including processes within the scientific community and interactions between scientific communities and other actors. Members of a scientific community are defined here as those who are principally involved as professional actors in such activities as pre-analytic theorizing, problem identification, hypothesis formulation and testing through various designs and procedures (such as mathematical modeling, experimentation or field study), data collection, analysis and data processing and critical validation through peer review and publication, i.e., activities commonly viewed as core practices of scientists.

     Intellectual investment in these activities by individual scientists is driven in part by human motivations such as curiosity and the pleasure of puzzle-solving but also by the structure of professional incentives that encourages-even demands-that scientists pay closer attention to obtaining the recognition of their work by peers in the scientific community rather than by other segments of society. However, scientific institutions cannot ignore the preoccupations and knowledge wielded by other actors (Girard and Navarette, 2005) and in other societal forums. This is particularly obvious in the case of agriculture; no matter the science involved in the origin and initial development of an idea, to be effective it has to become an applied science with potential for wide impact whose results are visible to all in the form of changes in agricultural landscapes. Thus it is unsurprising that opinions and drivers outside the domain of science itself condition science for agriculture. This tension between the incentives faced by individual scientists and the societal demands placed on scientific institutions in agriculture has been growing in recent decades, posing a strong challenge for the governance of scientific institutions (Lubchenco, 1998).

2.1.3.1 Cultures of science

Agricultural science processes in our period have been associated with the cultures of thought distinguished by two intellectual domains known respectively as "positivist realism" and "constructivism". The positivist realist understanding of modern science as a neutral, universal, and value- free explanatory system has dominated the processes of scientific inquiry in agriculture for the period under review. The basic assumptions are that reality exists independently of the human observer (realism), and can be described and explained in its basic constitution (positivism).This mind set is legitimate for the work that professional scientists do and enables transparent and rigorous tests of truth to guide their work. However, others (Kuhn, 1970; Prigogine and Stengers, 1979; Bookchin, 1990; Latour, 2004) have found this scheme problematic for explaining causality in their own disciplines for a number of reasons: it appears to exclude the qualitative (even if quantitative) ambiguous and highly contextualized interpretations that human subjects

give to the meaning of reality and it does not allow sufficiently for the unpredictability of the social effects of any intervention nor for the reflexive nature of social interactions (the object of enquiry never stabilizes; learning that something has happened changes decisions about what actions to take, in an unending dance of co-causality). This difference in legitimate perspective provides a partial explanation of why "the history" of the last sixty or so years cannot stabilize around a single authoritative causal interpretation of what has happened.

     For scientists working within positive realist traditions the locus of scientific knowledge generation is largely confined to public and private universities, independent science institutions and laboratories and to an increasing extent corporate research and development (R&D) facilities. These offer the conditions for highly specialized expertise to be applied to study of immutable laws governing phenomena that allow for prediction and control. Technology is conceived in this logic as applied science, i.e., as a design solution developed by experts removed from the site of application. The main task of the agricultural sciences in this perspective thus becomes that of developing the best technical solutions to carefully described problems (Gibbons et al., 1994; Röling, 2004). The problem description can and often does include scientists' understanding of environmental and social dimensions.

     The paradigm of positive realism has attracted largescale support for public and private science institutions as a way of thinking about and organizing innovation in tropical agriculture. It was harnessed to the expectation of maximizing yields and compensating for shortfalls in the quantity or quality of the biotic and abiotic factors of production by the provision of supplementary inputs, such as fertilizers and services to improve the productivity of labor and land. As such this paradigm lies at the heart of what is often called "productivism", a doctrine of agricultural modernization giving primary emphasis to increased productivity rather than the multifunctionality of agriculture or to the role of agriculture in rural development. It has constituted for much of the period under review a primary justification for science investments for development (Evenson et al., 1979).

     The dominance of this paradigm has had notable institutional consequences. University agricultural faculties progressively became divided into highly specialized departments. This split created "knowledge silos" that reflected the increasing specialization of scientific disciplines that reduced agriculture as an integrated practice into smaller and smaller fractions that largely excluded the human manager. This reductionism made it harder to mobilize multidisciplinary teams to address more complex problems (Bentley, 1994) and was consistent with the increasing specialization in modern farm sectors, developing countries and the social sciences.

     More inclusive and integrated science practices began to emerge from the 1970s onwards (Werge, 1978; Agarwal, 1979; Izuno, 1979; Biggs, 1980, 1982; Rhoades, 1982; Biggs, 1983). The drivers for this included the emergence of gender studies and women in agricultural development projects (Jiggins, 1984; Appleton, 1995; Doss and McDonald, 1999); the impact studies, analyses, and evaluations commissioned through the reporting cycles of the UN Human