HAND OUT 4 COURSE UNIT CARTOGRAPHIC DESIGN

TOPIC VISUAL VARIABLES

LECTURED BY HOD CARTOGRAPHY

 ❦
HOD CARTOGRAPHY

Cartographic communication. Further, such

Visual variables deconstruction reveals the value judgments and
power relationships implicit in the map design
process, allowing for critical analysis of the social
Overview of the visual variables construction and cultural negotiation of meaning
within the cartographic sign system.
Although the visual variables are an influential
The visual variables describe the graphic dimen- theoretical framework for cartographic design
sions across which a map or other visualization and research, the contents of visual variable
can be varied to encode information. The visual taxonomies vary considerably by scholar (Tyner
variables originally were described by French 2010). Figure 1 synthesizes several notable con-
cartographer and professor Jacques Bertin tributions within cartography into a set of twelve
(CE 1918–2010) in the 1967 book Semiologie visual variables, adapted from MacEachren et al.
Graphique. The English translation, Semiology (2012). While Figure 1 provides point sym-
of Graphics, was released in 1983 and today is bol examples for each of the visual variables,
recognized as a seminal theoretical work in both the visual variables can be used to encode
cartography and the broader field of information information about line and area features as
visualization. The visual variables were inspired well. Importantly, the twelve visual dimensions
by Bertin’s reading of semiotics, or the study of
❦
described here are only considered visual “vari-
sign systems. Semiotics seeks to understand how
ables” when functionally manipulated to encode
one object comes to stand for another object,
information. Each of these visual dimensions also
and therefore helps cartographers think about
can be manipulated as design embellishments
the way in which a map symbol (the sign vehicle
to improve the aesthetic quality of the map or
or signifier) representing a geographic phe-
visualization.
nomenon or process (the referent or signified)
Bertin (1967/1983) originally identified seven
comes to mean something (the interpretant) to
visual variables that can be manipulated to
the map user. French semiotics was developed
encode information.
in linguistics as a method for deconstructing a
language into its basic units in order to allow
for description and comparison across different 1. Location describes the position of the
language sign systems. Following this approach, map symbol relative to a coordinate frame.
Bertin’s visual variables describe the basic build- Location is considered an “indispensable”
ing blocks of a map or other visualization. By visual variable and takes visual primacy
reducing the map into its constituent graphic over the others. In cartography, loca-
elements, unclear or polysemic map symbols tion typically signifies the position of the
can be identified and redesigned, improving map symbol relative to a projected spatial
coordinate system, meaning that location
primarily is used to represent the spatial

❦
 ❦
HOD CARTOGRAPHY

VISUAL VA RIABLES

Ground
Figure

Location Y Y G G G

Size N Y G G G

Shape Y N G P P

Orientation Y Y G M M

Color hue Y Y G M M

Color value N Y P G M

❦
Texture Y Y G M M

Color saturation P G M

Arrangement M P P

Crispness P G P

Resolution P G P

Transparency M G P

visual variable variations Y = yes; N = no; G = good; M = marginal; P = poor; hatched = n/a

❦
 ❦
HOD CARTOGRAPHY

VI S UA L VA R I A B L E S

component of information in cartographic 4. Orientation describes the direction or

design. However, location can be used to rotation of the map symbol from “nor-
represent attribute information, such as the mal.” The normal orientation typically
use of perspective height in prism maps. is relative to the map’s neatline (either
Further, isoline maps manipulate the visual explicitly included or inferred by nega-
variable location, interpolating attribute tive space), but in some cases it can be
values to produce a spatial surface and relative to the projected spatial coordinate
system (e.g., relative to the graticule)
then weaving isolines through locations of
or another baseline. Multivariate glyph
equal attribute values within this surface.
symbols typically make use of orientation
In planimetric, 2-D representations, map
to differentiate among the represented
symbols near the optical center tend to attributes. Orientation also is manipulated
rise to figure, resulting in more immediate in flow maps to represent the directionality
visual interpretation by the map user and, of flow. Proximate clusters of map symbols
thus, landing them in a higher place of that have the same orientation (either
importance in the “visual hierarchy.” In aligned or misaligned to normal) tend to
contrast, map symbols near the periphery rise to figure as a single group. Otherwise,
of the page tend to recede to ground, individual map features that are misaligned
falling to the bottom of the visual hier- to normal tend to rise to figure.
archy. In oblique or 3-D representations, 5. Color hue describes the dominant wave-
map symbols that are perceived as nearer length of the map symbol on the visible
❦
or taller tend to rise to figure. portion of the electromagnetic spectrum
2. Size describes the amount of space (e.g., blue, green, red) and is one of three
occupied by the map symbol. Size is the visual variables associated with the per-
primary visual variable manipulated in pro- ception of “color.” Color theory occupies
portional symbol maps and value-by-area a substantial space in the cartographic
cartograms. Size also is manipulated in canon (Brewer 2005) and is particularly
relevant for choropleth mapping as well
flow maps that scale the thickness of the
as other forms of reference and thematic
flow lines to an attribute value. Larger map
maps that use colors to designate cate-
symbols tend to rise to figure.
gories or classes. A qualitative or spectral
3. Shape describes the external form (i.e., color scheme manipulates color hue while
the outline) of the sign vehicle. The visual controlling the other components of color;
variable shape is essential to the design as explained below, such color schemes
of qualitative point symbols commonly are appropriate for nominal choropleth
used in reference mapping. The shape of maps only. Red map symbols tend to rise
these map symbols can vary from highly to figure while blue map symbols tend to
abstract, such as circles, squares, or triangles, recede to ground.
to highly iconic, directly mimicking the 6. Color value describes the relative amount
referent represented by the map symbol. of energy emitted or reflected by the map
Map symbols that are more complex or symbol. Variation in color value results
less compact tend to rise to figure. in the perception of shading, or areas of

❦
 ❦
HOD CARTOGRAPHY

relative light (high emission or reflectance size of the texture units (related to the
of energy) and dark (low emission or visual variable size), and the density of the
reflectance of energy). Accordingly, color texture units (approaching the perceptual
value is sometimes referred to as “light- effect of shading associated with the visual
ness” in color theory. Manipulation of variable color value). Regarding the latter
color value is important for choropleth component, map symbols with a denser
maps depicting ordinal or numerical texture tend to rise to figure.
information. A sequential color scheme Bertin’s set of visual variables was extended
adjusts color value in one direction, some- by Morrison (1974) to include two addi-
times crossing over two or three different tional variables used in cartographic design.
color hues to improve discriminability. 8. Color saturation describes the spectral
A diverging color scheme adjusts color peakedness of the map symbol across the
value in two directions away from a critical visible spectrum, and is the third of three
midpoint, with each direction denoted visual variables associated with the per-
by color value changes within a differ- ception of color. Bold or saturated colors
ent color hue. A variation on a spectral emit or reflect energy in a concentrated
scheme with the green hues removed also band of the visible spectrum, whereas
is effective as a diverging color scheme. pastel or desaturated colors emit or reflect
The figure–ground relationship for color energy evenly across the visible spectrum.
value is relative to amount of light and From a design standpoint, therefore, color
❦ dark areas on the map. Map symbols that value can be conceptualized as the amount
are dark tend to rise to figure on maps of black in a map symbol while color
that are mostly light (e.g., with a white saturation can be conceptualized as the
background), while map symbols that amount of grey in a map symbol. Color
are light tend to rise to figure on maps saturation also is referred to as “chroma,”
that are mostly dark (e.g., with a black “intensity,” and “purity” in color theory.
background). Bold, saturated map symbols tend to rise
7. Texture describes the coarseness of the to figure while pastel, desaturated map
fill pattern within the map symbol. It symbols tend to recede to ground.
was once common to manipulate texture 9. Arrangement describes the layout of
in choropleth maps using halftone or graphic marks constituting a map symbol.
dithering techniques in order to mimic The visual variable arrangement varies
the appearance of shading; as described from regular (i.e., graphic marks are per-
above, contemporary choropleth maps fectly aligned in a grid-like structure)
instead manipulate the visual variables to irregular (i.e., graphic marks are ran-
associated with color due to advances domly placed or coalesce into clusters).
in modern printing and digital display Arrangement differs from the visual vari-
devices. Caivano (1990) describes tex- able texture in that all textures are assumed
ture as a higher-order visual dimension to be arranged regularly, regardless of
with three constituent components: the the initial direction, size, and density of
directionality of the texture units (related the texture. Dot density maps vary in
to the visual variable orientation), the both arrangement and texture density, a

❦
 ❦
HOD CARTOGRAPHY

higher-order visual dimension described in the linework. Map symbols that are
as “numerousness” by Nelsen (2000). Map depicted in a relatively high level of detail
symbols with irregular, and particularly tend to rise to figure.
clustered, arrangements tend to rise to 12. Transparency describes the amount of
figure. graphic blending between a map symbol
Finally, MacEachren (1995) identified and the background or underlying map
three additional visual variables whose symbols. MacEachren (1992) originally
manipulation is made easier through dig- referred to transparency as “fog” to suggest
ital production methods. MacEachren a partially opaque barrier impacting the
(1992) originally grouped these three clarity of the underlying map symbols.
visual variables under a single technique Transparency is the primary visual variable
called “focus” in context of uncertainty manipulated in value-by-alpha maps (the
visualization, but ultimately acknowledged “alpha” channel indicating transparency
each component as a visual variable given in computer graphics), an alternative to
the potential application to other forms of the value-by-area cartogram that visually
cartographic representation. equalizes enumeration units by adjusting
10. Crispness describes the sharpness of the their opacity rather than their size (Roth,
boundary of the map symbol. Crispness Woodruff, and Johnson 2010). Opaque
also is referred to as “depth-of-field” and map symbols tend to rise to figure.
“fuzziness” in information visualization.
❦ MacEachren et al. (2012) found that crisp-
ness was the most effective visual variable Perceptual basis of the visual
for representing uncertainty in the context variables
of point symbolization. Map symbols with and their syntactics for mapping
a crisp boundary tend to rise to figure
while map symbols with a fuzzy boundary
An important characteristic of the visual variables
tend to recede to ground.
is that they are processed pre-attentively, or in
11. Resolution describes the spatial precision
an immediate and preconceptual manner at the
at which the map symbol is displayed. The
sensory level of the human eye. Accordingly, the
visual variable resolution relates to the topic
direct translation of Bertin (1967/1983) is “reti-
of generalization in cartographic design,
nal” variable rather than “visual” variable. Thus,
which describes the meaningful removal of
the visual variables are “seen” perceptually rather
detail in the map design as the complexity
than “understood” cognitively. Principles of per-
of the real world is abstracted to fit the
ceptual psychology predict how each visual vari-
reduced scale of the map. Resolution as a
able is processed by the eye–brain system, and
visual variable leverages different levels of
therefore inform use of one visual variable over
abstraction to encode information, rather
others for cartographic design. Bertin postulated
than the typical purpose of generalization.
four such ways that visual perception may inform
In a raster depiction, resolution refers to
application of the visual variables in maps and
the coarseness of the grid size. In a vector
other visualizations, describing these four prop-
depiction, resolution refers to the amount
erties as “levels of organization.”
of detail (in terms of nodes and edges)
Bertin’s (1967/1983) first level of organization
is associative perception. With an associative

❦
 ❦
HOD CARTOGRAPHY

visual variable, variations in the visual dimension distribution of red symbols in Figure 2e than the
are perceived with equal weight, allowing for distribution of hexagonal symbols in Figure 2f,
the eye to perceive all map symbols with the despite the pair of figures encoding the same
same variation as a group (i.e., as associated). For information. Bertin believed shape to be the
instance, the eye is not drawn to one color hue only visual variable that was not selective. This
over the other in Figure 2a or one shape over property makes shape useful when each map
another in Figure 2b, allowing for the perception symbol should be interpreted individually, such
of all map symbols having the same color hue as in the context of large qualitative point symbol
or shape as an associated group; as a result, the sets. However, shape is not useful when scanning
eye is likely to see a series of horizontal rows for broad patterns across all map symbols, an
in Figures 2a and 2b. Because no variation important goal of most thematic mapping.
dominates, an associative visual variables allows Bertin’s (1967/1983) third level of organiza-
the eye to attend visually to other visual variables tion is ordered perception, while his fourth level
that also might vary in the visual scene (see the of organization is quantitative perception. Varia-
later discussion on visual variable conjunctions tions in ordered visual variables are perceived as
and bivariate mapping). Bertin believed location, ranked, with the eye pre-attentively interpreting
shape, orientation, color hue, and texture to be one variation as “more” or “less” than another
associative visual variables. With a dissociative variation. For example, the green symbols in
visual variable, one variation dominates visual Figure 2e are not perceived as “more” than
perception, with the eye drawn to this variation
the purple symbols, just different. In contrast,
❦ over others. In Figure 2c, the eye is drawn to
the darker symbols in Figure 2g are perceived as
the darker color values due to the contrasting
“more” than the lighter symbols, given the white
white background, while in Figure 2d, the eye
background. Bertin believed location, size, color
is drawn to the larger sizes. As a result, the
value, and texture to be ordered visual variables;
eye perceives a vertical gradient in Figures 2c
and 2d, rather than a set of horizontal rows as MacEachren (1995) later argued that color sat-
in Figures 2a and 2b, despite all four figures uration, crispness, resolution, and transparency
encoding the same information. Variation in a also are strongly ordered visual variables and that
dissociative visual variable inhibits attention to texture is only marginally ordered. Quantitative
other visual variables that may vary in the visual perception extends ordered perception, allowing
scene. Bertin believed size and color value to be for the estimation of numerical values from
dissociative visual variables. variations in quantitative visual variables. For
Bertin’s (1967/1983) second level of organi- example, the darker symbols in Figure 2g are
zation is selective perception. With a selective perceived as “more” than lighter symbols, but it
visual variable, the eye is able to focus individu- is difficult to estimate how much more without
ally (i.e., attend selectively) upon each variation use of a legend. In contrast, it is possible to
of the visual variable across the visual scene and estimate how much “more” the larger symbols
ignore the other variations. In other words, it is in Figure 2h are than the smaller symbols. Bertin
relatively easy to isolate visually the distribution believed quantitative perception to be restricted
of a particular category of map symbol across the to location and size only.
map when symbolized using a selective visual Importantly, Bertin’s (1967/1983) third
variable. For example, it is easier to see the and fourth levels of organization inform

❦
 ❦
HOD CARTOGRAPHY

(a) Color hue (b) Shape

=Associative =Associative
=Selective ≠Selective
≠Ordered ≠Ordered
≠Quantitative ≠Quantitative

(c) Color value (d) Size

≠Associative ≠Associative
=Selective =Selective
=Ordered =Ordered
≠Quantitative =Quantitative

❦
(e) Color hue (f) Shape
=Associative =Associative
=Selective ≠Selective
≠Ordered ≠Ordered
≠Quantitative ≠Quantitative

(g) Color value (h) Size

≠Associative ≠Associative
=Selective =Selective
=Ordered =Ordered
≠Quantitative =Quantitative

❦
 ❦
HOD CARTOGRAPHY

visual variable syntac- tics. In semiotics, different symbol dimensionalities. The images in
syntactics describes the relationship of sign Figure 3 encode the same pair of attributes, but
vehicles to one another. In cartography, Figure 3c represents them using a conjunction
syntactics prescribes the use of a visual that is homogeneous, Figures 3a, 3b, and 3d
variable given the level of measurement of the using a conjunction that is heterogeneous by
attribute information. Unordered visual visual variable, and Figure 3a using a conjunction
variables – such as color hue, orientation, and that is heterogeneous by symbol dimensionality.
shape – are appropriate for encoding nominal As with individual visual variables, perceptual
information. Visual variables that are ordered psychology informs the syntactics of visual vari-
but not quantitative – such as color value, able conjunctions. The most promising work
color saturation, crispness, resolution, and trans- to date is based on the sensory characteristic of
parency – are appropriate for encoding ordinal selective attention, or the ability to attend to
information. Finally, visual variables that are only one visual variable while ignoring others.
quantitative – such as location and size – are Visual variable conjunctions exhibit one of four
appropriate for encoding numerical information conditions of selectivity, each of which is appro-
but also can be applied for ordinal and nomi- priate for a different mapping context (Nelson
nal information given their visual dominance. 2000).
Figure 1 reproduces MacEachren’s visual vari- First, a separable conjunction describes a
able syntactics, designating each visual variable bivariate map or visualization in which selec-
as good, marginal, or poor for nominal, ordinal, tive attention of both attribute encodings is
and numerical levels of measurement. uninhibited. With a separable conjunction, the
❦
distribution of each attribute (“X” and “Y”)
can be “seen” without one restricting the other.
Visual variable conjunctions Separable conjunctions are produced when using
and bivariate mapping the visual variable shape, given its status as the
only nonselective visual variable. Conjunctions
Maps or other visualizations can make use of a that are heterogeneous by symbol dimensionality
conjunction of two visual variables. Conjunc- also tend to be separable, meaning that a thematic
tions can be applied for redundant symbolization, combination of choropleth or isoline (i.e., areas)
strengthening the graphic encoding of one with dot density or proportional symbol (i.e.,
attribute, or for representing multiple attributes points) will result in a separable conjunction. In
in a bivariate display. For bivariate maps, a con- Figure 3a, it is easy to attend to the distribution
junction can be homogeneous, using the same of either attribute (“X” and “Y”) but relatively
visual variable in two different ways, or het- difficult to determine their spatial correlation
erogeneous, using two different visual variables (“+”). A separable conjunction also is created
to represent the pair of attributes. In cartog- when combining size and color value on the
raphy, this distinction also relates to the map same symbol dimension, as both visual variables
symbol’s dimensionality (point, line, polygon, are dissociative; thus, shaded cartograms and
and volume), with a homogenous conjunction shaded proportional symbol maps are considered
manipulating the visual variables at the same separable. Elmer (2013) recommends use of
symbol dimensionality and a heterogeneous a separable conjunction when mapping two
conjunction manipulating the visual variables at independent variables that have incongruous

❦
 ❦
HOD CARTOGRAPHY

(a) Separable (b) Integral

• Independent attributes • Dependent attributes

• Incongruous scales

Choropleth +
graduated symbol Bivariate choropleth

Strong Strong

Moderate Moderate

Weak Weak
Size

Unknown/ Unknown/
variable variable

Color value

❦
(c) Configural (d) Asymmetrical
• Independent attributes • Attributes of different
• Congruous scales importance

Split graduated
symbol Value-by-alpha

Strong Strong
Transparency

Moderate Moderate

Weak Weak
Size

Unknown/ Unknown/
variable variable

Size Color hue

❦
 ❦
HOD CARTOGRAPHY

scales (e.g., different units of measure, different but there is also a visual cue aiding interpreta-
methods of normalization, different classification tion of the correlation between the attributes.
breaks), as there is no assumed correlation and Homogenous conjunctions that make use of
the user is forced to attend to each attribute scale split symbols are likely to be configural, as they
individually. exhibit a gestalt dimension of being “in-phase”
The conceptual opposite of a separable con- (i.e., having the same value for each half of
junction is an integral conjunction. With integral the symbol) or “out-of-phase”. In Figure 3c, it
conjunctions, selective attention is possible on is possible to attend to either side of the split
an emergent – or gestalt – visual dimension graduated symbol (“X” and “Y”) but also easy
(“+”) but selective attention of each original to interpret where both attributes are high (e.g.,
attribute encodings (“X” and “Y”) is inhibited. Utah) or low (e.g., Arizona) together, as the
In this situation, it is easy to “see” where the two halves reform the original circular shape.
two attributes are the same or different on their Elmer (2013) recommends use of a configural
respective scales, but it is difficult to attend to conjunction when mapping two independent
each attribute individually. Recommended color attributes that have congruous scales, as the
condition of being “in-phase” implies the same
schemes for bivariate choropleth maps make use
attribute value on the same attribute scale.
of an integral conjunction, with the arrangement
Finally, an asymmetrical conjunction describes
of colors producing an emergent dimension of
a situation in which a nonlogical gestalt
increased color value as both attributes increase.
dimension emerges in addition to the posi-
In Figure 3b, it is easy to determine where both
❦
tive correlation, as is the case with integral and
attributes are high or low in tandem by attending configural conjunctions. Here, the nonlogical
to changes in color value (“+”). However, it is emergent dimension (−) inhibits the ability
difficult to attend to only one attribute at a time; to “see” one of the two attributes (“X” or
for instance, both New Mexico and Oregon “Y”), producing an imbalanced visual effect
have the same value in the “X” variable, but in which the reader tends to interpret one
this is not easily interpreted due to the integral attribute over the other. Asymmetrical conjunc-
conjunction. Elmer (2013) recommends use tions tend to be produced when using size or
of an integral conjunction when the corre- color value with other visual variables, given
lation between dependent attributes is more their status as dissociative visual variables. An
important than the attributes themselves, as the asymmetrical conjunction also is produced with
gestalt dimension will focus the user upon this a value-by-alpha map (Figure 3d), with the
correlation. attribute encoded using color (“X”) visually
Separability and integrality represent ends of a equalized by the attribute encoded using trans-
continuum of conjunctive selectivity, with dif- parency (“Y”). Elmer (2013) recommends using
ferent conjunctions falling somewhere upon this an asymmetrical conjunction when one attribute
continuum in terms of visual strength. A con- (i.e., the variable of interest) is more important
figural conjunction describes a situation falling than the other (i.e., the equalizing variable).
in the middle of this continuum, where a gestalt The visual variables remain a central frame-
dimension exists (“+”) but the original attribute work for empirical research on cartographic
encodings are not fully inhibited (“X” and “Y”). design, informing both the experimental tri-
Here, it is possible to attend to each attribute als and controls. A similar approach rooted in

❦
❦
HOD CARTOGRAPHY

11