Research paper for PSY 506 Fall 2004 (Return to classweb page)
Stephen F. Austin State University
Studying visual variability discrimination allows researchers andpractitioners an opportunity to research how humans and animals (e.g.pigeons and primates) discriminate and perceive visual variability intheir environment. Visual variability discrimination is the abilityto recognize quantitative and qualitative differences between shapes,forms, and patterns. Practitioners are able to benefit from thisline of research. For instance, marketing firms are very interestedin finding ways to expand their sales record. Based on visualvariability discrimination research, Hoch, Bradlow, and Wansink, B.(1999) determined that if stores have a randomly organized display ofmerchandize, not only will the customer perceive there to be morevariety, but there is a strong chance that they will be a repeatcustomer.
Building architectural designers and city planning committees alsobenefit from visual variability research. Building designers shouldconsider that there is a positive relationship between variabilityand preference when they are in their development phase (Stamps,2002). Specifically, buildings should be more than a simple squarebox with windows, there should be creative architectural designs.Local and large corporations can also learn from this line ofresearch. For instance, companies that expand their line of computerprograms are more apt to adjust to changing technology. Furthermore,companies that have a diversified workplace are more able to reachcreative solutions than a homogeneous workforce (Young and Wasserman,2003).
As the aforementioned applications of visual variabilitydiscrimination were examples of how specifically this line ofresearch can help practitioners, visual variability research isgenerally conducted by research psychologists in laboratories withpigeons or human participants. Before any visual variability testingcan be conducted participants must receive training on how to respondto a visual display with differing degrees of variability.
Visual Variability Discrimination Training and Measurements
Many psychologists in this line of research train participants fordiscrimination between same-different visual displays. Same displaysinvolve the same icon repeated to complete the visual array, whereasdifferent displays involve non-duplicated icons. If people are usedin the study they receive either auditory or visual feedbackdependent on if they correctly respond to the trial. If pigeons areused they receive food reinforcement for correct responses, e.g. whena pigeon pecks the "same" button for a same display. Both types ofreinforcement allow the participants to improve their performance anddemonstrate competency in the visual variability discrimination taskbefore moving on to the testing phase.
Once the participants are in the testing phase the researchersmeasure their response to variability in the visual displays.Although variability is generally measured as a continuous variable,visual displays with icons form a categorical measure of variability.Entropy is used to measure variability on a quantitative andcategorical variable. Low variability is often referenced as lessthan 2.0 on an entropy scale, whereas, high variability is greaterthan 2.0. Young and Wasserman (2003) report that a person'sdiscrimination ability in visual display tasks is almost entirelyinfluenced by the level of entropy (variability) introduced in thedisplay. The remainder of this paper will explain current researchthat has been conducted on entropy, multidimensional entropy, andother research branches evaluating visual variability discrimination.
Entropy Research
Young and Wasserman (2003) found that as the level of entropy(variability) is reduced, participants require proportionately lessreaction time to respond to the visual display. However, if thenumber of icons in the visual display is reduced (e.g. from 9 to 4)and simultaneously entropy (variability) is increased, pigeonsperform at a declined level of accuracy and need more time torespond. Therefore, much research has focused on entropy because ithas been found to serve as a primary predictor of performance onvisual variability discrimination tasks.
In another study conducted by Young and Wasserman (2002) it wasdetermined through a variation of the same-different discriminationtask that humans and pigeons find it much easier to discriminatebetween uniformity and all levels of diversity. Their study alsodemonstrated that it is difficult for people to distinguish betweendisplays that include all different icons and those displays that aremixed with a combination of same and different icons.
Entropy research has also shown that humans are divided in howthey respond to a same-different visual display task. In a studyconducted by Young and Wasserman (2003), 55% of the participantsemployed relative entropy, while the other 45% used absolute entropy.Participants that use absolute entropy allow the number of icons inthe display and the variability of icons to affect their performanceon the discrimination task. These individuals generally respond thata different display is "different" when the number of icons isincreased and the "same" when the number of icons is reduced. Thesechanges in response due to number of icons in the visual display areonly seen for the different displays, not for the same displays.These results suggest that individuals that use absolute entropyperceive more variability (entropy) if the number of icons isincreased, regardless of actual variability.
Relative entropy, on the other hand, is not affected by the numberof icons in the visual display. The participants that use relativeentropy allow extra time and effort in calculating, first, theabsolute entropy, and second, the maximum possible absolute entropyof the array. In other words, they consider the number of icons inthe display and the icon variability, before deciding whether thedisplay is "same" or "different." Young and Wasserman (2003) foundthat participants that used relative entropy performed at a higherlevel of control and with a greater accuracy in distinguishingbetween high and low variability categories, than the participantsthat used absolute entropy.
Future research has yet to determine why some participants putforth the extra effort by using relative entropy. Young and Wasserman(2003) have eliminated GPA, gender, age, handedness, and ACT score aspredictors of using relative and absolute entropy. Additionally, itis questionable which strategy, relative or absolute entropy, wouldbe more difficult for participants to adapt to using in visualvariability discrimination testing.
Multidimensional Entropy
The studies discussed thus far have assessed items in asame-different discrimination task that either fall into the samecategory or the different category. In a study conducted by Young,Ellefson, and Wasserman (2003) other variables, besidessame-different icons were considered in the analysis, e.g.orientation, size, shape, and brightness. In order for these othervariables to be considered in the calculation of variability, a morecomplex measure of entropy must be employed. Young, et al. (2003)compares the Positional Entropy Model and the Finding DifferencesModel on their capacity to predict performance given these newvariables.
The Positional Entropy Model states that if duplicated icons areclustered together, humans and pigeons will be more likely toperceive the display as "same" than when the duplicated icons on thevisual display are less clustered and more widely distributed.Conversely, the Finding Differences Model incorporates the effect oflocalized differences through the use of a scale measuring theinfluence of distance on similarity. Specifically, the FindingDifferences Model pinpoints the icons on the visual display thatresult in the most activation, because they stand out and aredifferent from the neighboring icons.
Young, Ellefson, and Wasserman (2003) determined that both thePositional Entropy Model and the Finding Differences Model are goodmeasures of the effects on a pigeon's visual variabilitydiscrimination caused by the following manipulations: distance andclustering of similar items, increasing stimulus variability or mixedarrays, and the number of items in the visual display on same anddifferent arrays. However, only the Finding Differences test canaccount for the effects on variability discrimination caused by thedegree of similarity of each dimension related to size, shape, andorientation.
These findings support the idea that the Finding DifferencesModel, specifically the use of localized differences, better predictspigeons' performance on visual variability discrimination tasks(Young, et al. 2003). Future analysis should assess whether theFinding Differences model or some other tool predicts for humans'performance on visual discrimination tasks. Temporal distance shouldalso be considered in future studies of the Finding Differences Model(Young, et al. 2003). For instance, will the temporal presentation oficons in a visual display affect how humans and pigeons perform on avisual variability task?
Organization and Attention Research
Although entropy is the predominant predictor of variabilitydiscrimination it is not the only variable. Organization of icons ona visual display also affects responses on same-differentdiscrimination tasks. Wasserman, Young, and Nolan (2000) found thatif duplicated icons on a visual display are organized together,participants are less likely to report a "different" response thanwhen the icons are randomly placed on the visual display. Morespecifically, when the duplicated icons are organized in the samehalf of a visual display, the display will receive a lower percentageof "different" responses than any other organized pattern orrandomized visual display. On the other hand, randomized displaysreceive a higher percentage of "different" responses, than all othervisual displays, except the diagonal displays. The diagonal displaysare layouts that organize the similar icons into a diagonal line thatcrosses the visual display. The authors do not provide an explanationfor this finding. However, this last finding supports Hoch et al.'s(1999) report that random assortments of store merchandize offer theperception of greater variety and spur repeat customers.
Conclusion
Perception researchers have made many advances in discovering howpeople and animals process visual variability information. Not onlyhave researchers determined a quantifiable measure of variability(entropy), they have also determined that humans respond differentlyto visual variability tasks and that organization of icons affectshumans' perception of variability. Visual variability discriminationresearchers have clear directions for future research and manyresearch aspirations yet to achieve.
Hoch, S. J., Bradlow, E. T., & Wansink, B. (1999). The varietyof an assortment. Marketing Science, 18, (4), 527-546.
Stamps III, A. E. (2002). Entropy, visual diversity, andpreference. Journal of General Psychology, 129, (3), 300-320.
Wasserman, E. A., Young, M. E., & Nolan, B. C. (2000). Displayvariability and spatial organization as contributors to the pigeon'sdiscrimination of complex visual stimuli. Journal of ExperimentalPsychology: Animal Behavior Processes, 26, (2), 133-143.
Young, M. E., Ellefson, M. R., & Wasserman, E. A. (2003).Toward a theory of variability discrimination: Finding differences.Behavioural Processes, 62, (1), 145-155.
Young, M. E. & Wasserman, E. A. (2002). Detecting variability:What's so special about uniformity? Journal of ExperimentalPsychology, 131, 131-143.
Young, M. E. & Wasserman, E. A. (2003). Visual variabilitydiscrimination. In S. Soraci Jr. & K. Murata-Soraci (Eds.),Visual information processing (pp. 171-197). Westport, CT: Praeger.