THE THEORY OF COGNITIVE FIT - Human-Computer Interaction and Management Information Systems: Foundations - page 140

better supported with tables and high complexity problems with graphs. Note that these findings
are not necessarily consistent with all the tenets of cognitive fit.
To fully address cognitive fit, more finely grained analyses need to be conducted. The way in
which the researchers present their findings does not allow us to go beyond their own statements
because (1) they do not report analyses of pairwise comparisons; and (2) although they present graphs
of their interactions, they do so with results averaged over the third factor, which are the low and high
complexity conditions that we would like to analyze further. The researchers go on to state that: “Over
some intermediate area, performance differences between tables and graphs are minimal.” This state-
ment echoes some of those in Vessey (1994). However, because the authors had a stronger concept of
task complexity than of cognitive fit, they tended to overlook the role of task type.
The second study in this category is that of Speier and Morris (2003) who conducted an exper-
iment to assess the effects of task complexity (problem-solving task) on the task of finding a home
using an online system that had either a visual or a text-based query interface (problem represen-
tation). Task complexity was varied based on the number of alternatives in the low and high com-
plexity conditions: five viable homes in the low complexity condition and approximately 200 in
the high complexity condition.
The task is a multi-criteria decision making task that is essentially symbolic in nature. Each
interface supported an elimination-by-aspects decision-making strategy, which requires multiple
comparisons of alternatives, each of which is based on values of a single attribute. Although the
authors acknowledge cognitive fit as the foundational theory supporting their work, they do not
use cognitive fit to develop their hypotheses. Hence the role of the theory of cognitive fit in the
paper is unclear. For example, they base their arguments for hypothesizing that the text-based
interface will be better than the visual interface on an idiosyncrasy of the interface: The latter does
not facilitate ready comparison of “a small number of feasible solutions simultaneously.” They
hypothesized that the text-based interface would be more accurate than the visual interface on rel-
atively simple tasks, while the visual interface would be more accurate than the text-based inter-
face on more complex tasks.
In referring to decision-making time, the researchers stated their hypotheses based on the corre-
spondence of the task to “reality,” with reality undefined and therefore impossible to measure. They
argued that: “For most tasks, text-based characters on a screen are abstractions that result in only a
loose correspondence between the system and reality, while visual interfaces provide a more direct
mapping.” They further state that the low complexity task has a greater correspondence with reality
when supported by the text-based interface and that the high complexity task has a greater corre-
spondence to reality when supported by the visual interface. Based on their reasoning, the text-based
interface should result in faster decision making on simple tasks, while the visual interface should
be quicker for complex tasks. Note that the theory of cognitive fit, while having nothing to do with
arguments based on “reality,” leads to similar conclusions, albeit for different reasons.
The researchers found that, as hypothesized, the text-based interface resulted in greater accu-
racy when task complexity was low, while the visual interface resulted in greater accuracy when
task complexity was high. The findings for time, on the other hand, were contrary to expectations;
decision time was lower for the visual interface on the low complexity task and lower for the text-
based interface on the high complexity task. The authors offer no explanation for the fact that all
five of their findings with respect to time were in the opposite direction to those hypothesized.
The findings of this study support the theory of cognitive fit for accuracy on complex symbolic
tasks. Tables result in more accurate performance on simple tasks, those in which the number of
alternatives is small, while graphs result in more accurate performance when the number of alter-
natives is large. Because symbolic processing becomes increasingly difficult as the number of