Comparison of computer-simlated drafting with experiment shows that an Irregularity mechanism is present at the spinning frame but that its effects differ from those of some commonly proposed mechanisms. The strength of the mechanism Is shown to be a nuch more important determinant of yarn evenness than the evenness of the roving. This leads to the prediction, for example, that five gillings, rather than three, after combing should have a negligible effect on yam evenness if the saoie linear density of roving is spun.