Early studies of kiln structures suggested that magnetic refraction could be a major source of error in archaeomagnetic studies of large, strongly magnetized structures. Recent studies of the directions of magnetization of samples from four British and two Cretan kilns and furnaces do not show clear evidence for such effects, although only two show the level of scatter predicted from an assessment of the errors in orientation and measurement. There is also some evidence for systematic distortion of either declination or inclination around the circumference of some of these structures. These systematic deviations do not appear to be related to the intensity of magnetization, and the magnitude of scatter tends to decrease with increasing intensity of magnetization of the structure. The magnitude of scatter appears to be largest where the magnetic materials are inhomogeneous, but it is not clear how this can account for systematic variations. Evaluation of both earlier and newer observations indicates that floors are not necessarily the optimum parts of a structure for sampling, but instead walls can be considered at least as reliable if not more so. In both cases, differential subsistence may be the main source of error. It is recommended that, until more is known about the cause of scatter, sampling should incorporate both floors and walls, and that the azimuth of the sampling relative to the structure should be noted.