Powdered samples are assayed for potassium by the use of a Geiger counter designed for liquid samples. With precautions, the statistical accuracy expected in considerations of counting random events is achieved satisfactorily. A pure potassium compound is employed as an "absolute" standard of potassium content, obviating the use of calibrating solutions or synthetic standard mixtures. Full advantage of this is taken by employing a density correction, so that the use of an effectively "infinite" sample thickness, and hence large amounts of sample, is avoided. Unexpectedly, interference was experienced from uranium present in the phosphatic constituents of fertilizer samples tested. It was found possible to correct for this, and obtain satisfactory agreement with chemical determinations. The manipulation of powders instead of preparing solutions has advantages in rapidity for routine commercial or geological examination of many samples. Accuracy and sensitivity depend upon the usual statistical considerations encountered in radioactivity measurements. Good sensitivity and accuracy are attained without long counting periods, and it is considered that further developments are likely.