In this article, we examine the effect of sample size on spatial data quality in participatory mapping for assessing past land use. Using a map-based questionnaire, we capture changes in the extent and location of arable farmland 20 years ago. We then compare the results of the participatory mapping with reference data from the literature and accordingly calculate quality measures based on notions of correctness and completeness. In our data, correctness is high only for areas marked by many participants; the completeness metric increases as a function of the total area delineated by the participants. These data quality metrics then are analysed for subsamples, which are computed through leave-p-out jackknifing from our sample. This cross-validation addresses the question as to whether our data would be considered correct and complete having had included fewer participants. We present evidence that a small number of participants – in our case, less than 10 – already yield high-quality data. Further, we can demonstrate that aggregating the data from those participants with highest individual data quality values does not generate the best data quality as a group. This work provides a contribution to the use of participatory mapping for gathering correct and complete data.