Our hypothesis about the role of context in judgments predicts a statistical
interaction effect in each of the panels of Table 3: the difference between the
valuations of the myeloma and coral reefs problems is expected to be larger when
these items appear in the second position than in the first. The rationale for this
prediction is that the difference between the categories of ecological and human
problems is only salient when the issues are directly compared, not when they are
valued in isolation. The predicted interaction is highly significant p-.001. in each
of the four panels of Table 3.
The context effect observed in SWTP is especially noteworthy, because the
linguistic convention that allows words such as ‘important’ or ‘satisfying’ to be
understood in a relative sense does not apply to the dollar scale. To appreciate the
difference between scales that allow relativity and scales that do not, consider the
questions: ‘‘What is the size of an eagle, in meters?’’, ‘‘What is the size of a
subcompact, in meters?’’ Of course, there is no reason to expect any effect of
category on answers to this question. A context effect on a size judgment expressed
in absolute units indicates a visual illusion}a change in the underlying perception,
not in the language used to describe it. By the same logic, the finding of a context
effect on a dollar measure implies that the evaluation itself, not only the expression
of it, is altered by the comparison.