In the use of smoothing methods in data analysis, an important question is
which observed features are "really there," as opposed to being spurious sa
mpling artifacts. An approach is described based on scale-space ideas origi
nally developed in the computer vision literature. Assessment of SIgnifican
t ZERo crossings of derivatives results in the SiZer map, a graphical devic
e for display of significance of features with respect to both location and
scale. Here "scale" means "level of resolution"; that is, "bandwidth."