Conclusions


Appelogo

The Effects of Congruency Between Structural & Contextual Dominance in Image Processing




Conclusions

Response to the Specific Experimental Hypotheses

[H1]The first experimental hypothesis posed was that congruency between dominant structural elements and dominant contextual elements would result in greater attention to these congruent elements than in non-congruent situations. The statement, in terms of this experiment, looks specifically at the contextual dominance generated in each of the Congruency data groups. This dominance represents not only a contextual effect, but also the contribution of a structural effect as well. Because of the combined effect, it is called a congruency effect. In the Non-Congruent data group the dominance primarily represents the contribution of a contextual effect. In 69% of the cases, a focus on the contextually dominant element of the image was stronger when that element was also structurally dominant A Congruency Effect was particularly evident with the moderately complex images, where the average dominance was 34% higher than that in the Non-Congruent condition. These results lead us to a conditional acceptance of the first hypothesis. If the structural dominance is strong, then congruency between structurally dominant elements and contextually dominant elements will result in greater attention to these elements than in less-congruent situations.

One image was exemplary in demonstrating Congruency Effects -- image #8. Reviewing the structure of this image and context will clarify the types of "strengths" referred to above. The word "sky", heard in one group's contextual phrase referred to three sectors, A, B, & C, which were clearly related to the upper third of the image. The other group heard the word "base", which refers to the lower third of the image. The action verb, "points," which was heard by the Congruent data group, coincides with the visual convention of an arrow (which a pyramid shares in shape) that reinforces the congruency between contextual cue and image structure. The fact that the object was an isosceles triangle rather than an equilateral one increased the "pointing-up" effect even more (DB(CG-S) =89%). The Non-Congruent data group was directed to the base of the triangle by the contextual cue, but they had to overcome the strong image structure "pointing-up" to attend to the base. A very small portion of their dominance was devoted to the base area of the image (DH(NCG-S)=29%). This image produced the greatest Congruency Effect. It also demonstrated a strong Structural Effect evident in the inflation of sector B in the Non-Congruent data group (DB(NCG-S)=43%). There is little or no explanation for such dominant fixations to occur in this group except for the fact that the image structure was so strong as to draw fixations into sector B.

[go back to the BEGINNING of THIS document]

[H2]The second experimental hypothesis stated that, as complexity of an image increases, the structural dominance will decrease. Across the 15 images there was a continual trend downward in uniformity of agreement of structural dominance by the Experts. These results in uniformity, and loss of agreement on structural dominance, support the second hypothesis. Specifically it provides evidence that, at least with these images, as elements of equal or greater structural value are added to an image, the structural dominance declines in strength.

[H3]The third hypothesis was whether or not structural dominance is inversely proportional to the contextual effect; or, as structural dominance decreases, does the contextual effect have a greater impact on the attention patterns of the observer? With the information just presented on uniformity, if the hypothesis is true, then one would look for an increase in the contextual effect from the more complex images where uniformity was low, as was the structural dominance.

Referring to the graph in Figure 8, this hypothesis is supported by the Non-Congruent data group's performance exceeding that of the Congruent data group for images 13 through 15. As was mentioned in the detailed description of the complex images, the structural dominance was so illusive for the Experts that the primary dominance they indicated for the complex image set never matched that of the Free-Viewing data group. This success rate may be compared to a match of 60% in the moderately complex image set (images 6 through 10).

A case in point is that of image #14. One of the Experts described the way he thought that persons viewed a photograph was spatially (e.g. the same way one would if he were to be walking in that space). In other words, one would look at the first thing one would come to, then the second, then the elements surrounding one's immediate space, then beyond (in the case of image #14 this would produce an H, E, B progression of D ). Another Expert suggested that, since everything was arranged symbolically in a hierarchy from God, to Christ, to the sheep, and finally down to common man at the altar level, that the viewing patterns would coincide with that order (or a B, E, H progression of D ).

In fact, the viewing was more determined by the context presented by the verbal cue. The progression of D was H, E, B, D for the group which received a contextual cue relating to the alters at the bottom and center of the image, and B, E, A, H for the group which heard reference to the illustration in the half dome. This supports the third hypothesis that, as the structural dominance decreases, the context becomes more important as to where one focuses attention.

This finding would be supportive of the majority of research to date in the area of complexity, which states that simpler visuals are more predictable and memorable (Loftus, 1972; Luder & Barber, 1984; Friedman & Liebelt, 1981; Palmer, 1975; Heuvelman, 1987). New information is offered from this experiment regarding the relationship of Expert predictions, Congruency and Free-Viewing situations in terms of complexity. These variables have not been compared in the past.

[go back to the BEGINNING of THIS document]

The Concepts of Structural and Contextual Dominance

A major contribution of this study is its identification of Structural and Contextual dominance as factors which define Congruency, accompanied by a methodology for measuring them. Many studies have identified the structure of an image as being a variable (Heuvelman, 1987; Marschalek, 1986; Koroscik, 1984; Nodine, 1982; Molner, 1981 Nesbit, 1978; Dwyer, 1972), but none have defined it in terms of dominance. Many of these same researchers have presented a verbal or written phrase as a contextual setting to provide semantic encoding of the image, but none have compared contextual and structural dominance directly.

To the visual designer, the structural aspects are of utmost importance, since these are the elements that are being manipulated. The sad part is that messages are designed without attention to the structural dominance, or visual design, and because of their incongruity with the message, are ineffective. Research is done without concern to this very strong interacting variable, and inconclusive results are the outcome. The sooner that structural dominance takes its place beside contextual dominance as an equal contributor to the message, the sooner we will begin unraveling many of our communication issues involving images.

[go back to the BEGINNING of THIS document]

The Concepts of Congruency and Complexity

The finding in this study that Congruency Effects were strongest in the moderate complexity range supports Heuvelman's and Koroscik's finding that congruency can be higher with more abstract images. This appears particularly true when familiar visual conventions are used at this level of complexity. This study only partially supports other findings from these researchers that realistic or highly complex images distract to such a degree that congruency cannot be achieved. This study supports these views only when the stimulus consists of a weak contextual dominance coupled with a strong structural dominance. In these situations the dominance of the image takes over, and the contextual thrust is lost. However, this study also demonstrates that, especially in complex images, if the contextual dominance is strong and if the visual referents that are elements of the weaker structural dominance are also to by found within the image, then a contextual effect will also be strong, because of the contribution of a small congruency effect.

[go back to the BEGINNING of THIS document]

Summary

This study illustrates that when the structural dominance of an image is strong, an individual viewing this image will initially attend to the elements contributing to structural strength rather than to those which are contextual. The intent of most educators, in the presentation of an image, is mostly contextual. If a learner who rates low in knowledge of, and motivation toward, the specific information being presented, and if the image used is strongly structured in favor of a different context than that which the instructor intends, then the chances are high that the learner will not process the image in a desired manner. If this lack of congruency continues during a presentation, confusion on the part of the learner will be the most likely outcome. The converse of this situation is the desirable state of affairs. Instructional designers must gain adequate respect for the structural dominance issues, as the majority already have for the contextual dominance ones. The chances that learners encountering instructional messages will find the pathway to knowledge must greatly be enhanced. Most instructional designers do not have the skills necessary to analyze the structural components of an image to determine the structural dominance. Graphic designers, and a newly forming role called an interface designer (mainly involved with electronic media production), do have these skills, but most often are not adequately informed in the subject matter or instructional strategies to determine the appropriate contextual dominance. A team effort is necessary to assure that both aspects of the instructional development are adequately pursued. The inherent problem is that one person talks context while the other talks structure, and no progress is made. This study opens the door for communication between these two important people utilizing the rubric of Dominance. If both the instructional designer and the interface designer respond to each others' products in terms of the "dominance perceived," not only will the dialog be more fruitful, but the learner (who will be the ultimate "perceiver") will also be brought into the process. Placement of the message along a quantitative to qualitative continuum becomes a negotiation item for the "team." The interface designer will inherently be an advocate for more qualitative messages at one end of the continuum, while the instructional designer most likely will support more quantitative solutions at the other. Villemain (1966) supports the need for both in his statement that, "Qualitative mediations are instrumental to focally cognitive operations at one end of the spectrum, while on the other they become focal with cognitive elements assuming the instrumental role."

Major work to be done in research is in the field of art, not to quantify it and make it rigid, but instead to understand it adequately enough to incorporate it into the message design. The extent that anyone creating a message desires a level of perfection in communication, is the degree to which one embraces a desire to deal with the total continuum from information to art. Let John Dewey's challenge be our motivation:

As long as art is the beauty parlor of civilization, neither art nor civilization is secure... the ideal human community is dependent upon its esthetic component. Rather than envisioning art as an "efflorescence" , it is a condition of the realization of democracy, conceived as an ideal that lends a distinctive character to all aspects of life. If such a view is tenable, then significant revisions in democratic educational and social theory are in order. (Dewey, 1934)

[go back to the BEGINNING of THIS document] .. OR [go back to the MAIN MENU]



© 1996 Robert L. Appelman, PhD.


http://silver.ucs.indiana.edu/~appelman/D_FIVE.html





last updated: 8/24/96