Several cassava mosaic begomoviruses occur in the region, including African cassava mosaic virus, East African cassava mosaic virus and the recombinant Uganda variant, and mixed infections produce the most severe disease (Patil and Fauquet, 2009). Plants infected through the cutting show symptoms from the first leaves and yield least. Plants infected later by whiteflies show symptoms only on leaves formed after infection, and the later the infection the smaller the loss. Field studies in West Africa showed that most spread into a new planting came from sources outside it, with incidence highest on the edges facing the prevailing wind (Fargette et al., 1985). Epidemiological modelling that included vector dynamics indicated that the benefit of removing infected plants depends on how much inoculum arrives from outside the field (Holt et al., 1997). Both findings point to the neighbourhood, and not the single plot, as the scale at which infection pressure is set.

Thresh and Cooter (2005) reviewed the strategies available and grouped them under two headings: host resistance and phytosanitation, the latter meaning the use of virus-free cuttings and the removal, or roguing, of diseased plants. They noted that phytosanitation had seldom been adopted by farmers. The reasons are understandable. Pulling out a growing plant is a visible loss set against an invisible gain. Healthy cuttings are hard to recognise once stems have been cut and bundled. And a household that does everything correctly still receives viruliferous whiteflies from adjacent fields. The comprehensive reviews of Legg et al. (2015) and of the regional pandemics (Legg et al., 2011) reach a similar conclusion: sustained control requires action on planting material at a scale above the farm.

Cassava cuttings move mainly through gifts and exchange among relatives and neighbours, a pattern typical of the local seed systems described by Almekinders et al. (1994). The health of a village's planting material is in that sense a shared resource, which each household can improve or degrade. Ostrom (1990) identified the conditions under which communities manage shared resources successfully, among them clear boundaries, rules made by those who must follow them, monitoring by members, and graduated responses to rule-breaking. Group-based learning of the farmer field school type offers a tested way of building the diagnostic skills such management needs (van den Berg and Jiggins, 2007). The closest published precedent is the work on community phytosanitation against cassava brown streak disease in Tanzania, in which whole communities removed existing cassava and replanted with clean material (Legg et al., 2017).

The group first decides which fields are included. Because whiteflies move between adjacent plantings, the unit should be a continuous block and should take in as many of the cassava growers within it as possible; a patchwork of participating and non-participating plots defeats the purpose. The group then adopts a short set of rules covering where cuttings may come from, when roguing is done, and what happens when a member does not comply. Rules are written by the members with technical advice, not handed down.

Before the first planting under the scheme, and each season thereafter, trained members walk every field and record the proportion of plants with symptoms, the severity of symptoms on the usual five-point scale, and whether each infected plant was infected through the cutting or by whiteflies, which is judged from whether the lowest leaves show symptoms. Adult whiteflies are counted on the top leaves of a sample of plants. These are the methods standardised for regional surveys by Sseruwagi et al. (2004), and using them allows a village's figures to be compared with district and national data.

The group obtains its starting material from a source able to certify its health, such as a national research station or a registered seed entrepreneur. Varieties are chosen by the members from those with resistance, with attention to the taste, processing and maturity traits that decide whether a variety will be kept. The material is planted in a shared multiplication plot, sited away from old cassava where possible, and inspected closely. Stems from this plot supply members in the next season.

In the first months after planting, members remove any plant showing mosaic symptoms and fill the gap with a healthy cutting. Early roguing takes out the cutting-infected plants, which are the main sources of virus for whiteflies inside the block, at the stage when replacement is still worthwhile. Roguing is done on agreed days by members working together across all fields. Joint work makes the practice visible, spreads the labour, and removes the suspicion that one household is bearing a cost others avoid.

Shortly before harvest, while leaves are still on the plants, members mark vigorous symptomless plants as sources of stems for the next planting. Only marked plants supply cuttings for use or exchange within the group. This simple positive selection costs nothing and deals with the fact that a bundle of leafless stems gives no sign of its health.

At the end of the cycle the group meets to compare the season's survey with earlier ones, to hear from members who found the rules hard to follow, and to change what is not working. The comparison of cutting-borne with whitefly-borne infection is especially informative. A fall in the first shows that the seed measures are working; a persistently high level of the second shows that infection is arriving from outside and that the block needs to be enlarged or neighbouring villages drawn in.