DESIGNS THAT FIT - Human-Computer Interaction and Management Information Systems: Foundations - page 193

A review of empirical studies on cognitive fit (Vessey, this volume) concludes that the theory
is generally supported, and, furthermore, that the predicted relationship holds for a broader view
of fit than the original graph-table domain. Parallel research efforts (which do not use the term
“cognitive fit”) have also shown how incongruence between task demands and display hinders
performance, e.g., Jarvenpaa (1989). However, several factors limit the strength of the evidence.
First, there are no clearly defined measures of the main constructs (e.g., spatial-symbolic tasks) that
researchers can adopt unambiguously. Second, as a result of the first problem, there is no objective
measure of fit (beyond the binary fit relationship between pairs of spatial tasks—graphic display,
etc.). Third, the experimental designs employed in most of the reviewed experiments do not
observe directly the processes and representation assumed to be active in the user's mind. Finally,
the cost-benefit trade-off is hard to measure directly; thus, conflicting results between experiments
in terms of time spent and accuracy attained may be attributed to different priorities placed on
accuracy and time. For example, a task involving containment, in which one spatial object is con-
tained within another, is considered to be a spatial task in one study (Smelcer and Carmel, 1997)
and a symbolic task in another study (Dennis and Carte, 1998). Accordingly, the studies pre-
dicted, respectively, that a map or a table would provide a better fit with the perceptual processes
triggered by the map or analytic processes triggered by a table. There is however no direct obser-
vation of these processes and therefore no certainty that any one of these processes actually
occurred. This point opens cognitive fit to criticism that the theory is not falsifiable.
Developing measures of fit is key to the development of cognitive fit for both theoretical and
practical purposes. Measures are needed in order to test and compare applications of cognitive fit
and to generalize the model to other types of tasks. Goodhue (this volume) underscores the need
to develop measures for the higher-level concept of task-individual-technology fit. He raises two
alternative judgments of fit: the “objective” judgment of the experts and the subjective judgment
of the individual faced with the task. In both cases, judgment of fit must begin with a task analy-
sis that produces the true informational requirements. (In Figure 10.1, the task demands would
appear under the label Task.) However, having recognized the importance of defining task char-
acteristics as the basis for computing fit with individual and technology characteristics, the use-
fulness of the measure to designers depends on its granularity. For example, a questionnaire with
a user's evaluation of the task-technology fit based on perceptions of how well technology meets
one's job requirements (Goodhue, 1998) can help predict job performance or technology accept-
ance, but it can only provide general feedback to the designer on the technology fit, with no direct
input on the precise design. In contrast, a characterization of a task as one that requires certain
perceptual processes (Vessey and Galletta, 1991) provides specific guidance on the choice of dis-
play. In other words, it is not only that measures of fit must be developed, they must be developed
at the right level of granularity to be useful.
Norman (1991) advanced a very similar, yet broader, conceptualization of fit under the label of
“naturalness.” Naturalness is inversely related to the length of the translation between representa-
tions. The design implication is that the displayed representation “should allow the person to
work with exactly the information acceptable to the task: neither more nor less” (ibid., p. 29).
Unlike cognitive fit theory, this approach provides a basis for measures of fit, suggesting that fit
can be measured by the length of the description of the mapping, given a decision on the terms used
in the description. Norman proposes that researchers use psychological primitives to describe the
mappings and thereby build a common ground for comparing measures of fit. Following this rec-
ommendation, Kennedy et al. (1998) computed the number of elementary information processes
needed to perform a certain task based on quantitative displays of information. They used the
number of steps as a measure of cognitive fit, showing how people prefer displays with high,