Elementary students' reasoning about data modeling is explored by cond
ucting two design experiments. In the first design experiment, a class
of fifth-grade students worked in six different design teams to devel
op hypermedia documents about Colonial America. Students compared the
lifestyles of colonists with their own lifestyles. To this end, 10 ''d
ata analysts'' developed a survey, collected and coded data, and used
the dynamic notations of a computer-based tool, Tabletop (Hancock, Kap
ut, & Goldsmith, 1992), to develop and examine patterns of interest in
their data. Our general approach was to let the reasoning and thinkin
g displayed in one session with the data analysts provoke and steer th
e tasks and problems posed in the next session. Analysis of student co
nversations, including their dialogue with the teacher-researcher, ind
icated that the construction of data was an important preamble to desc
ription and inference. Moreover, students' ideas about many elements o
f data modeling were tied closely to forms of notation. In the second
design experiment, two children and their classroom teacher were consu
lted about the use of a simple randomization distribution to test hypo
theses about the nature of ESP. Here, experimentation afforded a frame
work for teaching about inference grounded by the creation of a random
ization distribution of the students' data. We conclude that data cons
truction and analysis provide an opportunity to involve students in th
e important enterprise of mathematical modeling.