A third way to explore stability is to investigate why many noisy processes tend
to produce mound-shaped distributions. Wilensky (1997) described a series of
interviews that he conducted with graduate students who were exploring this
question through computer simulations. We conducted a similar investigation with
fifth-grade students in an after-school program on data analysis. In analyzing a data
set on cats (from Rubin, Mokros, & Friel, 1996), students noticed that many
frequency distributions, like tail length and body weight, were mound shaped. As
part of exploring why this might be, students developed a list of factors that might
cause a cat’s tail to be longer or shorter. Their list included diet, being in an
accident, and length of father’s and mother’s tails. Using this list, we constructed a
spinner to determine the value of each factor for a particular cat’s tail. One student
might spin +2 inches for diet, +3 inches for mother’s contribution, –2 inches for an