Water yam (Dioscorea alata), an important root tuber crop in the tropics, is affected by anthracnose caused by Colletotrichum gloeosporioides. In order to reduce the impact of epidemics while reducing fungicide applications, the use of a combination of resistant yam cultivars and cultural practices is required. Among cultural practices, staking is traditionally used to grow yam vines in small fields, but is not in larger fields where mechanization is more widespread. In the literature, contradictory results have been obtained in the eighties by researchers when trying to assess the effect of staking on the dynamics of anthracnose. We compared the disease dynamics in both staked and unstaked plots under natural infestation in humid pedoclimatic conditions, by measuring weekly severity indices across a 12x8 grid made of 1 m² units spaced by 1.5 m. Additionally, leaf wetness sensors were placed at several locations in both plots, and leaf wetness duration (LWD) as well as other climatic and microclimatic variables (e.g. relative humidity, air and canopy temperatures, wind speed and direction) were monitored every 15 minutes for 41 days. Disease incidence (percentage of diseased units) could not be measured as all the units were diseased at the time of the first measurements; however severity indices suggest that the incidence was higher in the staked plot at a given time, and that the disease may have appeared earlier than in the unstaked plot. After fitting a 3-parameter logistic model to the data, we found that both the rate of increase of the disease and the time at the inflexion point were significantly different in both plots: the point of inflexion (time when 50% severity level is reached) occured on average at 154 ±1.9 and 165 ± 1.6 days after plantation for staked and unstaked plants, respectively), but the rate of increase was higher for unstaked than staked plots (11.22 ± 0.97 and 9.28 ± 0.76 dimensionless, respectively). Comparing the microclimate in staked and unstaked plots, cumulated LWDs were significantly longer in unstaked (491 ± 37 hours) than in staked plot (371 ± 70 hours). We observed that in the tropical conditions of the experiment, the high relative humidity of the air led to significant dew deposition during every clear night, and that dew was the main contributor to leaf wetness duration overall. Disease dynamics differences observed between staked and unstaked plots could not be readily explained by LWD differences only, since one would have expected faster disease development in the unstaked plot than in the staked plot whereas the opposite was observed. It is likely that in the humid pedoclimatic conditions of the experiment, leaf wetness duration was not a limiting factor for disease development, and that other factors at disease onset (such as the source and density of primary inoculum) play a critical role.