11.8 Aesthetics

The term aesthetics derives from a Greek word that means perception. The derivative modern meanings of beauty, taste, and artistic criteria arose in the 18th century. We have chosen the name aesthetics to describe the class of functions that turn theoretical graphs into perceivable graphics because of its original connotations and because the modern word perception is subjective rather than objective; perception refers to the perceiver rather than the object. Aesthetics turn graphs into graphics so that they are perceivable, but they are not the perceptions themselves. A modern psychologist would most likely call aesthetics in this sense stimuli, aspects, or features, but these words are less germane to our purpose.

Table 11.4 summarizes these aesthetic attributes. We have grouped these attributes in five categories: form, surface, motion, sound, and text. This is not intended to be an exhaustive list; other attributes, such as odor, can be devised. The color aesthetic has three components: hue, brightness, and saturation (other color components are possible). The texture aesthetic includes components of pattern, granularity, and orientation.

Table 11.4: Aesthetics

Form	Surface	Motion	Sound	Text
position	color	direction	tone	label
size	texture	speed	volume
shape	blur	acceleration	rhythm
rotation	transparency		voice

Seven of these attributes are derived from the visual variables of [6]: position (position), size (taille), shape (forme), orientation (orientation), brightness (valeur), color (couleur), and granularity (grain). Bertin's grain is often translated as texture, but he really means granularity (as in the granularity of a photograph). Granularity in this sense is also related to the spatial frequency of a texture.

These aesthetic attributes do not represent the aspects of perception investigated by psychologists. This lack of fit often underlies the difficulty graphic designers and computer specialists have in understanding psychological research relevant to graphics and the corresponding difficulty psychologists have with questions asked by designers. Furthermore, these attributes are not ones customarily used in computer graphics to create realistic scenes. They are not even sufficient for a semblance of realism. Notice, for example, that pattern, granularity, and orientation are not sufficient for representing most of the textures needed for representing real objects. Instead, these attributes are chosen in a tradeoff between the psychological dimensions they elicit and the types of routines that can be implemented in a rendering system. Specifically,

• An attribute must be capable of representing both continuous and categorical variables.
• When representing a continuous variable, an attribute must vary primarily on one psychophysical dimension. In order to use multidimensional attributes such as color, we must scale them on a single dimension such as hue or brightness, or compute linear or nonlinear combinations of these components to create a unidimensional scale.
• An attribute does not imply a linear perceptual scale. In fact, few aesthetic attributes scale linearly. Some attributes such as hue scale along curvilinear segments in two- or three-dimensional space. All linear scales are unidimensional but not all unidimensional scales are linear.
• A perceiver must be able to report a value of a variable relatively accurately and effortlessly when observing an instance of that attribute representing that variable.
• A perceiver must be able to report values on each of two variables relatively accurately upon observing a graphic instantiating two attributes. This task usually, but not necessarily, requires selective attention. This criterion probably isn't achievable for all of our attributes and may not even be achievable for any pair of them. But any attribute that is clearly non-separable with another should be rejected for our system. It is too much to expect, of course, that higher order interactions among attributes be non-existent. Much of the skill in graphic design is knowing what combinations of attributes to avoid.
• Each attribute must name a distinct feature in a rendering system. We cannot implement an attribute that does not uniquely refer to a drawable (or otherwise perceivable) feature. An attribute cannot be mapped to a miscellaneous collection of widgets or controls, for example.

We have attempted to classify aesthetics so that they are orthogonal in a design sense. One must not assume that this implies they are uncorrelated in our perceptual mechanisms, however. Orthogonalization in design means making every dimension of variation that is available to one object available to another. How these variations are perceived is another matter. Many aesthetic attributes, even ones such as size or position that are usually considered visual, need not be perceived visually. There is nothing in the definition of a graphic to limit it to vision. Provided we use devices other than computer screens and printers, we can develop graphical environments for non-sighted people or for those unable to attend to a visual channel because, perhaps, they are busy, restrained, or multiprocessing. Touch, hearing, even smell can be used to convey information with as much detail and sensitivity as can vision.

Every one of the figures in this chapter incorporates several aesthetics. Without aesthetic functions, they would not be visible. Consequently, we will not add a figure to illustrate other aesthetics, particularly since we are constrained in publishing format. Note, however, that in addition to using the position aesthetic function in every graphic, we have employed shape to differentiate symbols. Note, also, that position aesthetics are usually referenced by axes and shape and other aesthetics are usually referenced by legends.

Our discussion of the seven primary GOG components ends here. But there are several important topics remaining. We will first examine issues in graphics layout, and then conclude with a discussion of the relation between graphics algebra and statistical design models.