A visual variable, in cartographic design, graphic design, and data visualization, is an aspect of a graphical object that can visually differentiate it from other objects, and can be controlled during the design process. The concept was first systematized by Jacques Bertin, a French cartographer and graphic designer, and published in his 1967 book, Sémiologie Graphique. Bertin identified a basic set of these variables and provided guidance for their usage; the concept and the set of variables has since been expanded, especially in cartography, where it has become a core principle of education and practice.

Graphic techniques have been used in maps and statistical charts to represent non-visual information since the 17th Century, and information visualization blossomed in the 19th Century, highlighted by the work of William Playfair and Charles Joseph Minard. However, the direct study of this abstract use of graphical appearance began with the emergence of cartography as an academic research discipline in the mid-20th Century. In The Look of Maps (1952), often considered the genesis of American cartographic theory, Arthur H. Robinson discussed the role of size, shape, and color in establishing contrast in maps. At the same time in France, Jacques Bertin published an early version of his list of visual variables: shape, value, and "sparkling" (grain). Robinson, in his 1960 Elements of Cartography, which quickly became the dominant textbook on the subject, discussed size, shape, color, and pattern as the qualities of map symbols that establish contrast and represent geographic information.

Bertin was a cartographer at the École pratique des hautes études (EPHE) in Paris, where he created maps and graphics for faculty from various disciplines using a wide variety of data. Seeing recurring patterns, he created a system for symbolizing qualitative and quantitative information, apparently inspired by the sciences of semiotics, Human vision, and Gestalt psychology (it is sometimes hard to tell because his early works rarely cite any sources), culminating in Sémiologie Graphique. Despite having a background in cartography, and deriving many of his ideas by evaluating maps, he intended for Sémiologie Graphique to be applied to all forms of graphic design and information visualization. Soon the idea was gaining international acceptance; in 1974 Joel Morrison presented a very similar system in the context of cartographic generalization, citing neither Bertin nor Robinson but saying that it was a "traditional categorization," suggesting its widespread nature by that point. Several terms were proposed for this set of categories, including Bertin's "retinal variables" (used to distinguish them from his two spatial location variables), as well as "Graphic Variables," "Symbol Dimensions," and "Primary Graphic Elements," before eventually settling on "Visual Variables," as used almost universally (in English) today.

Bertin has largely been given credit for the system of visual variables; even though he was not the first to mention the idea, Sémiologie Graphique was the first systematic and theoretical treatment, and his overall approach to graphical symbolization is still in use today with only minor modifications. Despite the title of Bertin's work, it actually contained little reference to the scientific knowledge in the field of Semiotics or any other, and was primarily a practical summation of patterns he found in practice. The "truth" of the visual variables concept was largely established by its widespread and long-lasting acceptance. Thirty years later, MacEachren connected the scientific support for this and other aspects of cartographic design in How Maps Work, bringing together research in Semiotics (especially the Semiotic theory of Charles Sanders Peirce), Gestalt psychology, Human vision, and 40 years of cartographic research.

The earliest lists commonly suggested six variables: location size, shape, value, hue, orientation, and grain (pattern spacing). To this list, several additions have been suggested, with a few entering the canonical lists found in textbooks, while other suggestions have largely been dropped in cartography. With the rise of multimedia as a cartographic tool, analogous sets of non-visual communication variables have also been presented.

Starting with Robinson and Bertin, a core set of visual variables has become largely canonical, appearing in cartography and information visualization textbooks, and built into most design software in some form.

The size of a symbol is how much space it occupies. This commonly refers to the area of point symbols, and the thickness of line symbols. Size differences are relatively easy to recognize, making it a useful variable to convey information, such as a quantitative amount of something, or relative importance. Studies have shown that humans are better at judging relative differences in linear distance (e.g. one road being twice as thick as another) than relative differences in area (e.g., one circle having twice the area of another). Such estimations are the most accurate from squares. Area differences of circles are generally underestimated, but there is a large variation between people in ability to estimate two-dimensional size. Correctly estimating relative volume has proven even more difficult.

Because geographical features have an actual size on the Earth, this cannot always be controlled, and sometimes works against the wishes of a cartographer; for example, it can be difficult to make a world map in which Russia does not stand out. In a cartogram the size of features is purposefully distorted to represent a variable other than area.