The purpose of this thesis is to examine the problem of
employing unidimensional deterministic scaling in a multidimensional
probabilistic world. The best known unidimensional
deterministic scaling method is Guttman scaling. The scope
of inquiry is limited to scaling with four dichotomous scale
items. As a consequence of this examination, the paper
offers two contributions to the study of scaling. First, it
identifies two paradoxes of existing unidimensional deterministic
scaling methods. Second, it suggests a general framework,
called segment analysis, from which to approach
unidimensional deterministic scaling with dichotomous scale
items.