Food-for-work (FFW) projects face the challenge of addressing three kinds of objectives: to feed hungry people, to build public works where needed, and to be feasible for prompt project implementation. In the debate over how to target FFW to the poorest of the poor, the last two program objectives are often overlooked. This research examines FFW afforestation and erosion-control programs in central Tigray, Ethiopia, during 1992-95 in order to examine how these sometimes conflicting objectives were reconciled. The study decomposes the factors determining a household's FFW participation into three decision stages. First, at the regional level, project planners choose where to locate a FFW resource conservation project. Second, at the village level, a committee decides which villagers will be eligible to participate. Finally, the eligible households may decide whether and how much to participate. Using probit and truncated regression methods, the study finds that project implementation feasibility most influenced the probability that FFW projects would be available in the 25 villages surveyed. Among the 129 households in villages with FFW available, FFW eligibility was inversely related to household land area per capita and household size, results which are consistent with anti-poverty targeting. However, the model performed poorly at predicting non-eligibility for FFW, which suggests that anti-poverty targeting was not efficient. Among households eligible for FFW, those with greater resources (larger families that did not lease out land) tended to participate and supply more days of FFW labor than poorer households. The only households eligible for FFW which did not participate were unable (rather than unwilling) to do so, being comprised mostly of elderly women. Overall, anti-poverty targeting was sub-optimal but reasonable, given the feasibility constraint that these resource conservation projects to be sited in where labor and materials could be made available.