AbstractIf the three‐dimensional form of a hillslope is regarded as an infinite population of two‐dimensional slope profiles, values for attributes of the form of the hillslope can be determined by sampling and estimating the means of the attributes of the population of slope profiles. Twelve attributes are examined in seven first‐order drainage basins located on the South Downs in East Sussex to determine necessary sample sizes. Results show that if spatial randomness is assumed the size of sample required varies considerably from one attribute to another but is commonly very large. Of four alternatives to spatial randomness that are tested, consideration of both hillslope plan form and valley asymmetry are shown not to lead to reduction in sample size. However, results do suggest that maximum slope, percentage of profile length in the range 5° to<10°, and mean angular difference between adjacent slope segments may all be dependent upon distance of the slope profile down valley. Furthermore, initial tests of the effect of spatial autocorrelation suggest that for some attributes consideration needs to be given to this factor when determining sampling str