This study contains the results of an experiment in which signers were asked to judge the acceptability of a set of sign manipulations. Signs were recorded with variations in different categories, in the temporal and spatial dimension. Participants varied much in tolerance, i.e. in the percentage of movies they judged to be acceptable, but their rankings of the acceptability of sign manipulations correlated well. On the level of dimensionality, we found that temporal manipulations were highly acceptable while spatial (and spatiotemporal) manipulations were often judged unacceptable. Further division of the manipulations into categories, such as changes in hand orientation or movement direction, showed much variability and little regularity. The roles of phonology and iconicity in acceptability judgments were studied. Part of the variability in the acceptability of sign manipulations could be explained on the basis of the type of phonological error caused by each manipulation, and by considering a sign's iconicity and classifying whether manipulations are compatible or incompatible with that iconicity. Finally, human judgments were compared to acceptability ratings by three automatic sign recognizers for three of the signs. With some exceptions, the rankings of the acceptability of sign manipulations as calculated by these sign recognizers did not correlate well with the rankings by humans.