There is no one widely accepted method of managing international agricultural research and numerous different models exist. Here we review one in particular, referred to as the ‘institute without walls’, from the perspective of the UK Department for International Development's (DFID) Renewable Natural Resource (RNR) Research Strategy (1990–2006). We begin with a brief history of the RNR Research Strategy from 1990 to 2004. We then draw on nearly 15 years experience of managing one of the programmes within the RNR Research Strategy to assess critically the impact of externally and internally imposed organizational and management changes on the performance of the DFID Plant Sciences Programme (PSP). The current RNR Research Strategy (1995–2006), with its emphasis on demand-led research, has greatly increased the relevance and effectiveness of DFID's natural resources research. A comparison between the PSP in 2004 and the early 1990s inevitably concludes that the programme has been transformed: unlike in 1991, research is now firmly demand-driven, much is based in developing countries and farmers are benefiting from the research. Over time, the outputs of the long-term strategic research have been applied in practical plant breeding and participatory crop improvement programmes. Key to the success of the PSP has been the provision of continuous, long-term funding which has allowed projects time to develop and produce outputs of real value to end users. Alongside this, the ability of the PSP to build long-term, in-country partnerships has ensured the effective adoption of its research outputs. We conclude that the successes of the PSP have largely derived from (i) identification of research that is clearly demand driven, (ii) continuous long-term funding that has allowed research to move from the strategic to adaptive phase, (iii) continuity of management, and (iv) the flexibility to develop a wide range of partnerships, both in-country and overseas, based on their ability to deliver.

Three cultivars of Coffea arabica, Catuaí Vermelho IAC 81, Icatu Amarelo IAC 2944 and Obatã IAC 1669–20, were evaluated in relation to leaf gas exchange and potential photochemical efficiency of photosystem II under field conditions on clear and cloudy days in the wet season in southeast Brazil. Independent of levels of irradiance, leaf water potential (υleaf) values were always higher than the minimum required to affect daily net photosynthesis (PN). PN, stomatal conductance (gs), leaf transpiration (E) and the index of photochemical efficiency (Fv/Fm) declined on a clear day in all cultivars. The depression of leaf gas exchange and Fv/Fm (specially around midday) caused a strong decrease (about 70 %) in daily carbon gain on a clear day. Under cloudless conditions, gs and PN were correlated with the air vapour pressure deficit (VPDair), but not with photosynthetic photon flux density (PPFD) values. On a cloudy day, the daily carbon gain was barely limited by PPFD below 800 μmol m−2 s−1, the Fv/Fm values showed a slight decrease around midday, and gs and PN were positively correlated with PPFD but not with VPDair. By contrast, irrespective of the contrasting irradiance conditions during the day, PN and E were correlated with gs.

Over an eight-year period, harvesting methods based on simple mechanical aids (blade and shear) were evaluated against hand harvesting on mature morphologically contrasting tea clones in Southern Tanzania. The effects of shear step height (5–32 mm) and the harvest interval (1.8–4.2 phyllochrons) were also examined. Except in the year following pruning, large annual yields (5.7–7.9 t dry tea ha−1) were obtained by hand harvesting at intervals of two phyllochrons. For clones K35 (large shoots) and T207 (small shoots), the mean harvested shoot weights were equivalent to three unfurled leaves and a terminal bud. The proportions of broken shoots (40–48 %) and coarse material (4–6 %) were both relatively high. Using a blade resulted in similar yields to hand harvesting from K35 but larger yields from T207 (+13 %). The yield increase from clone T207 was associated with the harvest of more shoots and heavier shoots, smaller increases in canopy height, and a higher proportion (7–9 %) of coarse material compared to hand harvesting. On bushes, which had been harvested by hand for two years following pruning, using flat shears (no step) supported on the tea canopy resulted, over a three year period, in yields 8–14 % less than those obtained by hand harvesting and, for clone K35, a reduction in the leaf area index to below 5. The development of a larger leaf area index is made possible by adding a step to the shear. However, since annual yields were reduced by 40–50 kg ha−1 per mm increase in step height, the step should be the minimum necessary to maintain long-term bush productivity. As mean shoot weights following shear harvesting were about 13 % below those obtained by hand harvesting, there is scope, when using shears, to extend the harvest interval from 2 to 2.5 phyllochrons.

Genotype × environment interaction plays an important role in the identification of genotypes with high and stable yields. The environmental components of such interaction can be attributed to location (spatial) and year (temporal) factors. This study considers a specific aspect of the location factor, i.e. that locations not geographically close to each other are likely to vary in terms of the conditions that determine the effect of the year factor, even during the same calendar year. In these situations, the year effects are considered to be random variables within each location. This study addresses the question of whether genotype × location interactions are consistent (or repeatable) over time: a question which is crucial when defining the target environments in a plant breeding programme and when deciding the optimum environment for selection. Using three sets of data from field experiments, various computation methods were developed to estimate variance components; the methods and estimates are presented here. Using the bootstrapping procedure to sample one year from each location within a given calendar year, we obtained bootstrap replicated values of genotype × location interaction and their summary statistics, including the coefficient of variation (CV). We suggest that the CV can be used to measure the repeatability of genotype × location interactions. A Genstat program that can be used to perform the required computations is available from the authors upon request.

Water is perhaps the most important single factor that limits crop production in the semi-arid northeast of Nigeria. A four-year field experiment was therefore initiated in 1999 to evaluate the influence of land configuration practices with or without wood-shavings mulch on water conservation, yield and water use efficiency (WUE) of sorghum. The experimental treatments consisted of flat bed (FB), open ridging (OR), tied ridging (TR), FB + mulch (FBM), OR + mulch (ORM), and TR + mulch (TRM). Ridge heights were 15 to 20 cm and furrows were left open (for OR and ORM) or tied (for TR and TRM). Wood-shavings mulch was used at the rate of 5 t ha−1 in 1999, but this was increased to 10 t ha−1 in subsequent years to ensure adequate soil coverage. Differences in soil water storage at various sampling dates were significant only in some cases in each year, but trends were towards greater soil water storage in the mulched treatments than in the non-mulched treatments, irrespective of tillage method. Growth parameters (plant height and leaf area index) indicated significant differences between treatments on only some measurement dates in each year. Sorghum water use varied significantly between years and treatments. Seasonal water use was greater with FBM, ORM and TRM than with the FB treatment in all cropping seasons. Averaged over the four-year period, mean increases in grain yield relative to the FB treatment were 16 % for OR, 25 % for TR, 77 % for FBM, 50 % for ORM and 57 % for TRM. Pooled across the experimental years, the WUE (ET) of FB, OR, TR, FBM, ORM and TRM were 1.95, 2.12, 2.13, 2.74, 2.36 and 2.48 kg ha−1 mm−1 respectively. The corresponding WUE(R) values for these treatments were 1.26, 1.46, 1.56, 2.22, 1.88 and 1.97 kg ha−1 mm−1 respectively. It is concluded that combining the practice of flat bed cultivation with mulching may eliminate the need for ridging in increasing the productivity of sorghum grain in semi-arid northeast Nigeria.

This paper analyses the way the CGIAR system has incorporated social research in its agenda. Since 1995, the social science staff capacity in the CGIAR has decreased by 24 %, and the overall balance of social science research is still significantly tilted away from the core germplasm enhancement, production systems/natural resources management, and technology adoption work – the ‘bread and butter’ of technology generation and development effort – toward ex-ante and ex-post activities. Further, the bulk of the social science research has low social research content despite the significant expansion of the CGIAR initial goal of increasing the proverbial ‘pile of rice’ to poverty alleviation and sustainable food security. The paper concludes that a concerted effort is now required to mainstream social research in the CGIAR system, and this cannot occur without the full support of the CGIAR donors, the CGIAR senior managers, and the centre boards and executive staff.

Once in a while, every plant breeding project has to face a difficult situation: for example, objectives not reached, insufficient production of new genetic material, poor variety adoption and context change. In most cases, the problems are biological in nature, but they may also arise from poor situation analysis leading to inappropriate selection criteria or failing seed systems. We believe that plant breeders need special tools for analysing the human situations in which they are immersed. Such tools would assist them in identifying strategies which are favourable for one, several or all of the actors in a situation. Comprehensive sociology proposes a method derived from actionist analysis. This method relies on a situation description organized in a Panoramic or Pan Table (PT) and on empathetic induction by the breeder of the actors' positions. The PT facilitates the identification of the zones of potential conflict. This paper is methodological. We show how to use the PT method in a real situation for tactical or strategic purposes. In our example, the breeder has to elaborate a specification for a cotton breeding project in the case of a commodity chain that is co-ordinated by a national cotton company. After describing the PT method and its use, we discuss its limitations and the way human problems can be taken into account when elaborating strategies for action.