Assignment errors in histograms with equal class intervals of arbitrary width

It is shown that significant distortions can occur in histograms constructed with a class interval of arbitrary width, due to the inclusion (or omission) of readings in a class interval which do not (do) properly belong to that interval. Unless the class interval width is at least ten times greater than the reading interval the proportion of unrepresentative readings may exceed 10%. In any event caution is required in the interpretation of a histogram in which this effect is present.