Although the literature is replete with guidelines as to how soil and plant samples should be obtained, soil sampling remains a fragile link in the soil testing process, and interpretation of plant analytical data is often fraught with uncertainty. There is unquestionably a need to re‐examine currently accepted procedures and to identify aspects which might lead to an improvement in the quality of samples submitted to advisory laboratories. This paper discusses some of the weaknesses in soil sampling identified in Natal, South Africa, and an aspect of plant sampling usually overlooked, but deserving of much closer examination. None of the issues discussed are uniquely original, but all have had an impact on soil testing and plant analysis in South Africa and may well have application elsewhere. Aspects of soil sampling discussed include the sampling tool, the influence of subsoil properties on the value and meaning of topsoil analytical data, and heterogeneity in farmer fields and experimental plots. Choice of sampling tool is considered to be an issue of particular importance, since sample quality Is strongly related to the labour and time required to obtain samples and many of the tools popularly employed are not conducive to high intensity sampling. A sampler is described which has all the attributes of other acceptable equipment, but which is particularly easy to use. The importance of subsoil properties is evaluated by examining the influence of subsoil acidity, K distribution in the profile and vertical K movement on topsoil‐based fertilizer recommendations. Brief discussion is included on the effects of soil heterogeneity in farmer fields on sample integrity and of implement carry‐over of nutrients in long‐term calibration trials. Plant sampling is examined in the light of fortuitously acquired findings on the influence of soil moisture on nutrient uptake by maize. Data presented demonstrate that variations in soil water content several weeks prior to sampling may invalidate commonly accepted tissue content norms. These data might help to elucidate anomalous findings and encourage re‐examination of an issue considered to be fundamentally important to the diagnosis of plant nutritional status.