Tufte’s principles of

data visualization
Edward Tufte

 American Statistician
 A pioneer in the field of data
visualization
 Noted
 for his writing on Information Design.
 (also for his criticism of Power Point!)

 Main idea:
 Minimalism: Show as much data as
possible with as little ornamentation
He wrote, designed, and self-published 4 classic books on data
as Possible.
visualization.
Tufte’s Main DataViz Design Idea

Show as much data as possible with as little ornamentation as Possible

 Above all else, show the data
 Maximize the data-to-ink ratio
 Erase non-data-ink
 Erase redundant data-ink
 Revise and edit.

infovis-wiki.net/wiki/Tufte,_Edward
Tufte’s Principle

 Graphical Integrity
 Data-ink Ratio
 Chart Junk
 Data Density
 Small multiples

The Visual Display of Quantitative Information, 2nd Edition, 2001

Tufte’s Principle of Graphical
Integrity
 Tell the truth: Avoid distortion of data in visualization, whether
accidental or intentional.
Distortion of Data

https://en.wikipedia.org/wiki/Misleading_graph
Intentional Distortion of Data

ex: Steve Jobs, rather famously, once used a

distorted 3D bar chart to show Apple's share of the
US SmartPhone market.

See more from https://en.wikipedia.org/wiki/Misleading_graph

Tufte’s Principle of Graphical
Integrity
 Tell the truth: Avoid distortion of data in visualization, whether
accidental or intentional.
 The representation of numbers (as physically measured on the
chart/graph), should be directly proportional to the numerical quantities
represented.
Tufte’s Principle

 Tell the truth: Avoid distortion of numbers by graphic devices,

whether accidental or intentional. Show data variation in context,
and label them.

Dramatic viz
but misleading:
As a reader, you
simply look at
how the lines
have grown and
conclude that the
fuel economy is
not only
improving but
improving more
and more rapidly
as the lines get
bigger and bigger
faster.
This graphic was originally published by the NY Times. It tries to show the mandated fuel economy standards for autos set by the US Department of Transportation.
Data Integrity: Lie Factor

 Lie factor is a measure of misrepresentation

In Graph:
(5.3 – 0.6)/0.6=7.83
In Data:
(27.5 – 18)/18=0.53

Lie Factor:
7.83/0.53 = 14.77

https://infovis-wiki.net/wiki/Lie_Factor
Fuel Economy Standard
Truthful representation of the data
by Tufte
Data Integrity

Lie Factor: 1
Compute & Compare Lie Factor
Size of effect shown in graph: (15-5)/5 = 2
Compute&Compare Lie Factor
Size of effect shown in graph: (15-5)/5 = 2 Size of effect shown in graph: 10/40 = 0.25

Lie Factor =
2/0.25 = 8
Data Integrity: Lie Factor
Show data variation not design variation.

Note:
Area is less efficient for mapping data.

Lie Factor = 2.8

https://infovis-wiki.net/wiki/Lie_Factor
Data Integrity: Lie Factor
Show data variation not design variation.

Lie Factor = 2.8

https://infovis-wiki.net/wiki/Lie_Factor
Data Integrity: Lie Factor

The image for B is actually 9 times as large as A.

LieFactor = 9/3 = 3
Length to Area
Tufte's Principles of Graphical
Integrity (summary)
1. Representation of numbers should match the true proportions.
2. Labeling should be clear and detailed.
3. Design should not vary for some ulterior motive, show only data
variation ;-)
4. The number of dimensions represented should not be more than the
number of dimensions in the data.
5. Representations should not imply unintended context.
Tufte’s Principle: Data-Ink Ratio

 Maximize Data-Ink Ratio

“A large share of ink on a graphic should present data-information,
the ink changing as the data change. Data-ink is the non-erasable
core of a graphic, the non-redundant ink arranged in response to
variation in the numbers represented.” [Tufte]

https://infovis-wiki.net/wiki/Data-Ink_Ratio
Tufte’s Principle: Data-Ink Ratio

 Minimalism: Good graphics should include only data-Ink.

 Non-Data-Ink is to be deleted everywhere where possible. The reason
for this is to avoid drawing the attention of viewers of the data
presentation to irrelevant elements. The goal is to design a display with
the highest possible data-ink ratio (that is, as close to the total of 1.0),
without eliminating something that is necessary for effective
communication.

https://infovis-
wiki.net/wiki/Data
-Ink_Ratio
Terrible Data-Ink Ratio

 Analysis:
 The data-ink ratio is abysmally low.
 There are only 11 data points in this
entire graphic.

 the image contributes nothing

meaningful to the interpretation of
the data.
 the 2D data is made 3D
 perspective exaggerates the
decline.
 Exaggeration of 2.5 percentage
point decline over twenty years
Tufte's Principles Related to Data-
Ink Ratio
 Above all else show data.
 Maximize the data-ink ratio.
 Erase non-data-ink.
 Erase redundant data-ink.
 Revise and edit
High Data-Ink Ratio Example

An electroencephalogram: a graph that records the electrical activity from the brain, has a data-ink ratio of 1.

The Visual Display of Quantitative Information, 2nd Edition, 2001

Tufte’s Principle: Chart Junk

 Chartjunk – the excessive and unnecessary use of graphical effects

in graphs.
 Avoid chartjunk.

“The interior decoration of graphics generates a lot

of ink that does not tell the viewer anything new.
The purpose of decoration varies —
• to make the graphic appear more scientific and
precise,
• to enliven the display,
• to give the designer an opportunity to exercise
artistic skills.
• …
Regardless of its cause, it is all non-data-ink or
redundant data-ink, and it is often chartjunk”.
[Tufte]
https://en.wikipedia.org/wiki/Chartjunk
Chart Junk: Example

 3D Bar Chart:
 There is no information in the use of
the 3rd dimension which heavily
influences the pictures and very
much drives the attention towards
the depth of the bars and the
underlying grid. But this is not at all
the information which should be
transported.

https://infovis-wiki.net/wiki/Chart_Junk
Chart Junk: Example

 Confusing Graphics:
 The logos giving a hint what the
lines are dealing with are far more
than the minimum needed to
present the desired information.

https://en.wikipedia.org/wiki/Chartjunk
Tufte’s Principle:
• Data Density

 Maximize data density.

 The data density of a graph is the proportion of the total size of the
graph that is dedicated displaying data.
 Human eye registers 150 Mbits and the brain can process and
understand this data.
 We should take advantage of this high bandwidth to maximize
information transfer rate for our visualizations.
 discourage visualization that allocates only a smallish portion of the
display to data and the rest to widgets and other non-data part of the
visual.
Tufte’s Principle: Data Density

 Maximize data density.

 The data density of a graph is the proportion of the total size of the
graph that is dedicated displaying data.
High Data Density
Minard's map of Napoleon's disastrous Russian campaign of 1812.

A cartographic depiction of numerical data on a map of Napoleon's disastrous losses suffered during the Russian campaign of 1812. The illustration depicts
Napoleon's army departing the Polish-Russian border. The visiualization is notable for its representation in two dimensions of five types of data:
(i) the number of Napoleon's troops (thickness) (ii) distance (inset) (iii) temperature (bottom plot) (iv) direction of travel; and (v) location relative to specific
dates. [source: https://en.wikipedia.org/wiki/Charles_Joseph_Minard#Work]
This type of band graph for illustration of flows was later called a Sankey diagram.
Tufte’s Principle: Small multiples

 small multiples are a great tool to visualize large quantities of data

and with a high number of dimensions.
 series of the small similar graphs displaying different slices of the data set
repeated using the same scale and axes in one visual.
"Illustrations of postage-stamp size are indexed by category or
a label, sequenced over time like the frames of a movie, or
ordered by a quantitative variable not used in the single image
itself.“ [Tufte]

 Small multiples go by many names: Trellis Chart, Lattice Chart, Grid

Chart, and Panel Chart.
Small multiples: Example
Small multiples: Example
Public support for School
vouchers, broken down by
religion/ethnicity, income, and
state

https://www.juiceanalytics.com/writing/better-
know-visualization-small-multiples/
Sparklines

 Sparklines: data-intense, design-simple, word-sized graphics.

 Sparklines are small enough to be embedded in text,
 Example:

 “the sparkline as a kind of ‘word’ that conveys rich information

without breaking the flow of a sentence or paragraph made of
other ‘words’ both visual and conventional” [Tufte]

Wiki
Tufte’s Principle: Small multiples

 A sparkline is a very small line chart,

typically drawn without axes or
coordinates.
 It presents the general shape of the
variation (typically over time) in some
measurement in a simple and highly
condensed way.
 ex: temperature or stock market price,
 Several sparklines may be grouped
together as elements of a small
multiple.

Wiki
Tufte’s Principle: Small multiples

 A sparkline is a very small line chart,

typically drawn without axes or
coordinates.
 It presents the general shape of the
variation (typically over time) in some
measurement in a simple and highly
condensed way.
 ex: temperature or stock market price,
 Several sparklines may be grouped
together as elements of a small
multiple.

Wiki
Tufte’s Design Principles: Summary

 “Overload, clutter and confusion are not attributes of information,

they are failures of design” Edward Tufte.
 Any visual property that does not represent data (ex: Shade, label,
color, …) should be included to aid the visual perception, not hinder it.
 Visualization Ethics: Must avoid deception.

Note:
 Tufte’s principles are a good starting point for data visualization.
 They are not a religion.

https://blog.plotly.com/post/143370337552/above-
all-else-show-the-data
Data Visualization
PRINCIPLES
  Human Visual functions are
extremely fast and efficient
 Cognitive functions are much
Designing slower and less efficient.
Visualization  Designing visual functions that
should take advantage of the
strengths of visual functions and
help in cognitive function.
Visual Processing of
Information
Visual Pathway
 Signal from the eye is passed onto
the optic nerve, the primary set of
cells that transmit signal to the brain.
 At optic chiasm, the optic nerves from
each eye cross.
 The optic chiasm superimposes
information from both eyes together,
so that you only see one image.
 From there it passes through lateral
geniculate nucleus and finally
reaches the primary visual cortex, the
main area of the brain that interprets
visual data. This is where our mind
starts to recognize that the eyes are
seeing.

The primary visual cortex, often called V1, is

a structure that is essential to the conscious
processing of visual stimuli. https://www.brainfacts.org/Thinking-Sensing-and-
Behaving/Vision/2012/Vision-Processing-Information
LGN and Visual Cortex

 The image captured by

each eye is transmitted to the
brain by the optic nerve. This nerve
terminates on the cells of
the lateral geniculate nucleus, the
first relay in the brain's visual
pathways.
 The cells of the lateral geniculate
nucleus then project to their main
target, the primary visual cortex.
 It is in the primary visual cortex that
the brain begins to reconstitute
the image from the receptive
fields of the cells of the retina.

http://thebrain.mcgill.ca/
Stages of Visual Processing

 Transduction – changes light energy into signal

 Transmission – Neuro signal goes through low-
level processing and is transmitted to visual
cortex
 Selection – aspects selected of stimulus
 Organization – grouping of elements to form a
whole
 Interpretation – given meaning
Transduction

 The visible part of the Electromagnetic energy

(light) reaching the eye is transduced
(converted) to electrochemical energy by
receptor cells (Rods and Cones).
 Receptor cells respond to 380 nm (violet) to 760
nm (red) of the electromagnetic radiation
spectrum. This is the visible light spectrum to
which humans are able to respond.
Rods VS Cones

RODS CONES
 Black and white vision  Daytime and colour vision
 Can’t discriminate colours  Excellent visual acuity – sharp and
clean images
 Operate well in low level light – night
vision  Concentrated in the fovea
 Sensitive to brightness, darkness and  Not useful at night.
movement
 Outnumbered by rods 19:1
 Mainly located in outer part of retina
 Poor acuity (low sharpness and
focus)
 Excellent for peripheral vision rather
than direct
 19x the number of cones
Retinal Ganglion Cells

Wiki
Transmission

 Electrochemical signals from photoreceptors

undergo low-level processing in ganglion
receptive field,
 Processed signals are carried over by optic
nerves to the visual cortex.
Retinal Ganglion Cells

Wiki
Selection

 Process of discrimination and identification of

the FEATURES in the stimulus.

 Feature detectors are cells that are in the retina

and visual cortex, that respond to patterns, lines,
edges and angles.
Stages of Visual Perception

 Transduction – changes it so brain can

understand it
 Transmission – Neuro signal goes through low-
level processing and is transmitted to visual
cortex
 Selection – aspects selected of stimulus
 Organization – grouping of elements to form a
whole
 Interpretation – given meaning
Organization: Grouping

 This can only happen once the visual cortex has

received the neural impulses and features are detected.

 Feature elements are grouped to form a whole, using

perceptual principles that work like ‘rules’ of
organisation.

 These principles are called Gestalt principles.

Gestalt Principles

 Gestalt = German for “Pattern”, “form” or “shape”

 Gestalt principles outlines grouping laws that help us
 group individual elements of a visual stimulus into a complete form.
 understand how the wholes are structured and organized.
 and theorizes that human visual system perceives objects in the
simplest way possible.

Wiki
Law of prägnanz:
Fundamental Gestalt Principle
 We tend to order our experience in a manner that is
 regular,
 orderly,
 symmetrical, and
 simple.
 So Keep it simple. Arrange data logically wherever possible.

Wiki
1. Gestalt Law of Proximity

 “Things that are close together are perceptually

grouped together”.
 Also known as nearness. Elements nearby together form an
overall figure or pattern.
 Best way of emphasizing relationships between different
data entities is to place them in proximity in a display.
Proximity example

“Storytelling with Data” Cole Nussbaumer

Proximity example

Leveraging proximity in Creating Tables

“Storytelling with Data” Cole Nussbaumer

Proximity

 When reading a scatter

plot, we tend to group
data points that seem
near each other and
assume they share similar
characteristics.
 In the scatter plot on the
right we can easily define
two groups.

“Everything is related to everything else, but near things are more

related than distant things” Tobler’s first law of Geography
2. Gestalt Law of Similarity

 Elements with similar features (size, shape, color,

orientation) are grouped together.
 Similarity is stronger than proximity if the two are together
in a stimulus.
 Combined with proximity creates stronger visual
grouping.
 Rule: Related visual elements should look similar
Similarity Example

“Storytelling with Data” Cole Nussbaumer

Similarity

 When we use different

colors to identify series in
a chart we are applying
the law of similarity: we
group together things
that share the same
color, shape, direction.
3. Gestalt Law of Common Fate

 Regardless of how far apart the elements are placed or

how dissimilar they appear, if they are seen as moving or
changing together, they’ll be perceived as being
related.
Common Fate Example
4. Gestalt Law of Enclosure

 Elements that are physically enclosed together create a

visual grouping.
 Enclosure does not have to be strong. Background
shading can be enough.

“Storytelling with Data” Cole Nussbaumer

Enclosure Example

“Storytelling with Data” Cole Nussbaumer

Enclosure Example
Leveraging Enclosure to draw distinction
between different part of the data

“Storytelling with Data” Cole Nussbaumer

5. Gestalt Law of Closure

 Closure:
 We want to see the things to be simple and fit into the
constructs that are already in our head.
Fido?
5. Gestalt Law of Closure (contd.)

 We want to see the things to be simple and fit into the

constructs that are already in our head.
 We perceive an object as a whole, despite it being
actually incomplete.
 We group the individual elements to make ‘one’ by filling
in the missing contour lines so that it makes sense.
 Closed contours segment visual space. Organizing using
these segmented regions is stronger than proximity.
Closure Examples

“Storytelling with Data” Cole Nussbaumer

Closure Examples
Leveraging Closure we can strip the darker background,
heavy border the graph still appears a whole object and
the data stands out for us.

“Storytelling with Data” Cole Nussbaumer

6. Gestalt Law of Continuity

 Arrange objects in a line or simple curve to facilitate

grouping and comparison.
Continuity

 People are more likely to construct visual groupings out

of elements that are smooth and continuous, rather than
ones that contain abrupt changes in direction.
 Visual system wants to see continuous line/curve even
when they do not exist.
 Implies connectedness, which is stronger than proximity.
Continuity Example

“Storytelling with Data” Cole Nussbaumer

Continuity Example
Leveraging Continuity we can continue
stripping things out of our visualization. More
we strip away the more the data stands out.

“Storytelling with Data” Cole Nussbaumer

7. Gestalt Law of Connectivity

 Rule: use connections to show relations

 Has a stronger associative value than most other laws

 Not typically stronger than enclosure, but depends on

the thickness and darkness of enclosure.
Connectedness Example

“Storytelling with Data” Cole Nussbaumer

Connectedness Example
We Leverage Connectedness in drawing line
graphs, so that out eyes makes sense of a
whole lot of data.

“Storytelling with Data” Cole Nussbaumer

8. Gestalt Law of Figure/Ground
and Metastability
 We tend to separate the important aspects of the figure
from the surroundings (background). We focus or give
our attention to the figure.
 Real or imagined contour lines separate the figure and
ground
Metastability

• Metastability arises
because of the difficulties
with establishing contour
Love or
lines death
• Reversible figures change
the ownership of the
contour lines
Gestalt Law Figure/Ground and
Metastability
 We tend to separate the important aspects of the figure
from the surroundings (background). We focus or give
our attention to the figure.
 Real or imagined contour lines separate the figure and
ground
 Ensure there is enough contrast between your
foreground and background so that charts and graphs
are more legible.
Metastability

Do we have one bar chart,

or two bar charts?

https://emeeks.github.io/gestaltdataviz/section4.html
 Figure and Ground

Simply outlining one region

establishes it as Figure and
the other as Ground.
 Figure and Ground

Simply outlining one region

establishes it as Figure and
the other as Ground.
Figure and Ground

Avoid strong hue or highly

saturated color as ground.
Application of Gestalt Laws
Decluttering step by step

“Storytelling with Data” Cole Nussbaumer

Application of Gestalt Laws
Remove the chart border

“Storytelling with Data” Cole Nussbaumer

Application of Gestalt Laws
Remove Grid Lines

“Storytelling with Data” Cole Nussbaumer

Application of Gestalt Laws
Remove Data Markers

“Storytelling with Data” Cole Nussbaumer

Application of Gestalt Laws
Clean-up Axis Labels

“Storytelling with Data” Cole Nussbaumer

Application of Gestalt Laws
Label Data Directly (Proximity)

“Storytelling with Data” Cole Nussbaumer

Application of Gestalt Laws
Leverage Consistent Color (Similarity)

“Storytelling with Data” Cole Nussbaumer

Application of Gestalt Laws

“Storytelling with Data” Cole Nussbaumer

Summary: Gestalt Laws in Data
Visualization
 Law of Proximity: Know what your chart’s information priority is, and
then create groupings through proximity to support that priority.
 Law of Similarity: Use similar characteristics (color, size, shape, etc.) to
establish relationships and to encourage groupings of objects.
 Law of Closure: Close contour is perceived even if one does not exist.
So avoid clutter.
 Law of Continuity: Arrange objects in a line to facilitate grouping and
comparison.
 Law of Figure/Ground: Ensure there is enough contrast between your
foreground and background so that charts and graphs are more
legible.
 Law of Common Fate: Use direction and/or movement to establish or
negate relationships.
Optical Illusions
How do Optical Illusions work?

 A visual stimulus misleads our perception (or meaning-

making) of that stimulus.

 This happens because we APPLY perceptual

constancies to what we are seeing – they are our RULES.

 We make a false judgement because we misjudge

length, position, speed, direction or curvature.
Interpretation

 The ‘whole’ that has been created is given meaning.

 Memory is used to compare incoming information with

what is already known.

 This creates a perceptual hypothesis.

 Interpretation involves the brain using psychological

factors in order to make sense of what it is considering.
MULLER LYER Illusion

 There are 2 theories to

explain this illusion!
 The Muller Lyer illusion
consists of 2 lines that
are equal in length to
each other.
• One has fish-tails’
attached.
• One has ‘arrow-
heads’ attached

https://en.wikipedia.org/wiki/M%C3%BCller-Lyer_illusion
Theory of Perceptual Compromise

1. The length of the solid horizontal lines of each figure

are actually equal in length.
2. BUT the length of each WHOLE figure is not the same.
3. The lengths between the added tips (heads or tails) is
very different. We use closure to estimate the length
and compromise somewhere between the actual
length and the perceived length.
Diagram of Perceptual Compromise
Diagram of Apparent Distance
Theory
 We use our knowledge from depth
cues and size constancy.
 We associate the line with ‘fish-tails’
as being the inside corner of a room
in a building.
 We associate the arrow-heads as
the closest outside corner of a
building .
 BOTH lines cast the same size image
on the retina.
 As the fish-tails are perceived to be
further away (inside the building),
we apply size constancy and
interpret that the line must be
longer.
Size Illusion and Data Visualization

Possible size
misinterpretation
in bubble charts

https://visual.ly/blog/illusions-in-data-visualization/
Visual Attention
Visual Attention

“Storytelling with Data” Cole Nussbaumer

Color Contrast and Visual Attention

“Storytelling with Data” Cole Nussbaumer

Use Color contrast effectively
Color

 Color is important
 Why we need it?
 To represent data
 To bring the data layer to the front
 Confirm to design requirement
(recap) Effectiveness of Visual properties
for Different Types of Data
Quantitative Ordinal Nominal
 Position on the axis  Position on the axis  Position on the axis
 Length of line or object  Color Saturation  Color Hue
 Area of shape  Color Hue  Color Saturation
 Color Saturation  Shape
Color in plotting packages
Plotly Default Color choices
Plotly Default Color choices
Color Preference

 The tendency for an individual or a group to prefer some colors over

others. Say a favorite color.
 Infants begin showing a preference for color is at about 12 weeks old.
 Children prefer the colors red/pink and blue.

Wiki
 Color Preference (By Gender)
Most Favorite Color Male Female

Source:
http://munsell.com/color-
blog/best-color-infographics/
Data:
http://www.joehallock.com/edu/COM498/surveydata.html
 Color Preference (by gender)

 Blue reigns Supreme:

 Both males and females like
blue (35% female, >50% male).
 Why is Blue Popular?
 associated with clean water,
clear sky, authority, truth,
tranquility etc..
 a personal favorite among all
age groups and genders.

Source: http://munsell.com/color-blog/best-color-infographics/
Data: http://www.joehallock.com/edu/COM498/surveydata.html
 Color Preference (by gender)

 The Color Purple

 Most notable gender
difference
 23% female consider as most
favorite, no male chose it as his
favorite!
 Why?

Source:
http://munsell.com/color-
blog/best-color-infographics/
Data:
http://www.joehallock.com/edu/COM498/surveydata.html
 Color Preference (by gender)

 The Color Purple

favorite among
females
 Gender Stereotype?
 Parents commonly dress
their baby girls in pink and
their baby boys in blue.

Ref: Pretty in pink: The early development of gender-stereotyped colour

preferences
Article (PDF Available) in British Journal of Developmental
Psychology 29(Pt 3):656-67 · September 2011
https://www.nickkolenda.com/color-psychology/
 Color Preference

https://munsell.com/color-
blog/best-color-infographics/
 Color Preference: (By gender)
Least Favorite Color Male Female

• Male and female

participants disliked many of
the same color
• Orange and Brown are
the least favorite colors

Source:
http://munsell.com/color-
blog/best-color-infographics/
Data:
http://www.joehallock.com/edu/COM498/surveydata.html
 Color Preference By Gender

 Bright vs. soft colors

 Women prefer soft colors
 Men prefer bright ones
 Achromatic colors
 Men tolerate achromatic color more
than women
 Tints vs. Shades
 Women preferred tints more than
shades.
 Opposite for men
Source:
http://munsell.com/color-blog/best-color-
infographics/
Ref: 2007 Study by Hulbert and Ying
 Men keep it simple!

Men-Women
Color Names
What may be “purple” to men, is
“grape”, “eggplant” or “plum” or
any other fruit/flower to Women

http://munsell.com/color-blog/best-color-
infographics/

Data Source:
http://www.datapointed.net/visualization
s/color/men-women-color-names-d3/
Color Association
Findings from a global survey of 200K people
 Happy – Yellow  Powerful - Red (tomato)
 Pure – White  Dependable – Blue
 Good Luck – Green  High Quality – Gold
 Good-tasting - Red  Nausea - Muted Yellow
(tomato)  Deity – White
 Dignity - Dark Blue  Bad Luck – Black
 High Technology – Silver  Favorite Color – Blue
 Mourning – Black  Least Favorite Color - Dark
 Expensive – Gold Yellow
 Inexpensive – Brown

https://www.colorcom.com/global-color-survey
Color: Meaning and Association

 References:
 Color Psychology: https://www.nickkolenda.com/color-psychology/
 color-theory-for-designers-part-1-the-meaning-of-color:
https://www.smashingmagazine.com/2010/01/color-theory-for-
designers-part-1-the-meaning-of-color/
 Best color for infographics: https://munsell.com/color-blog/best-color-
infographics/
Warm and Cool Colors

 “Warm” colors are associated

with warm feeling: daylight or
sunset:
 hues from red through yellow,
browns and tans included
 "cool" colors associated with
cool feeling: water, green
grass, a gray or overcast day:
 hues from blue green through Wiki
blue violet, most grays included.

https://www.webopedia.com/TERM/C/cool_colors.html
Warm vs Cool Colors

 It is ironic that
 the hotter stars radiate blue (cool) light, and the cooler radiate
red (warm) light.

temperature of hot bodies

Wiki
Warm vs Cool Colors

 Warm colors
 tend to advance or appear more active in a visualization,
whereas cool color tend to recede.
 are said to stimulate the viewer, where as cool color tend
to calm and relax the viewer
 Children prefer cool colors over warm colors.

Wiki
Warm Color: Red

 Red is a very hot color. It’s associated with

 fire, violence, and warfare.
 love and passion. (In history), associated with both the Devil and
Cupid.
 anger
 importance (ex: red carpet treatment)
 danger (ex: stop light and signs, warning labels)
 Red can actually have a physical effect on people,
 raising blood pressure and respiration rates.
 (shown to) enhance human metabolism, too.

https://www.smashingmagazine.com/2010/01/color-
theory-for-designers-part-1-the-meaning-of-color/
Warm Color: Red

 Cultural differences:
 In some eastern culture,
 red is the color of prosperity and happiness, and is said to attract
good luck.
 red is worn by brides on their wedding days.
 In South Africa, however, red is the color of mourning.
 Red is also associated with communism.

https://www.smashingmagazine.com/2010/01/color-
theory-for-designers-part-1-the-meaning-of-color/
Warm Color: Red

 In design and visualization:

 red can be a powerful accent color.
 WARNING: can have an overwhelming effect if it’s used too much in
designs, especially in its purest form.
 a great color to use when power or passion want to be
portrayed in the design.
 brighter versions being more energetic and
 darker shades being more powerful and elegant.

https://www.smashingmagazine.com/2010/01/color-
theory-for-designers-part-1-the-meaning-of-color/
Warm Color: Orange

 Orange is a very vibrant and energetic color.

 Because it is associated with the changing seasons it can
represent change and movement in general.
 Because orange is associated with the fruit of the same
name, it can be associated with health and vitality.
 In visualization
 orange commands attention without being as overpowering as
red.
 often considered more friendly and inviting, and less in-your-
face.

https://www.smashingmagazine.com/2010/01/color-
theory-for-designers-part-1-the-meaning-of-color/
Warm Color: Yellow

 brightest and most energizing of the warm colors. It is

associated with
 happiness and sunshine.
 WARNING: can be associated with deceit and cowardice,
though (ex: in 10th century France, door of traitors and
criminals were painted yellow).
 hope, (in some countries yellow ribbons are displayed by
families who have loved ones at war).

https://www.smashingmagazine.com/2010/01/color-
theory-for-designers-part-1-the-meaning-of-color/
Warm Color: Yellow

 In design and visualization

 Bright yellow lends a sense of happiness and cheerfulness
 Softer yellows are commonly used as a gender-neutral
color for babies (rather than blue or pink) and young
children.
 Light yellows also give a more calm feeling of happiness
than bright yellows.
 Dark yellows and gold-hued yellows can sometimes look
antique and be used in designs where a sense of
permanence is desired.
https://www.smashingmagazine.com/2010/01/color-
theory-for-designers-part-1-the-meaning-of-color/
Cool Color: Blue

 Blue represent calmness and responsibility.

 Light blues can be refreshing and friendly.
 Dark blues are more strong and reliable.
 In design and visualization the exact shade of blue you select will
have a huge impact on how they are perceived.
 Light blues are often relaxed and calming.
 Bright blues can be energizing and refreshing.
 Dark blues, like navy, is used where strength and reliability are important.

https://www.smashingmagazine.com/2010/01/color-
theory-for-designers-part-1-the-meaning-of-color/
Cool Color: Green

 Green is a down to earth color.

 has many of the same calming attributes that blue has, but
it also incorporates some of the energy of yellow. It is
associated with
 new beginnings and growth
 renewal and abundance
 envy or jealousy, and lack of experience.

https://www.smashingmagazine.com/2010/01/color-
theory-for-designers-part-1-the-meaning-of-color/
Cool Color: Green

 Brighter greens are more energizing and vibrant,

 Olive greens are more representative of the natural
world.
 Dark greens are the most stable and representative of
affluence.
 In design and visualization
 balancing and harmonizing effect, and is very stable.
 appropriate to relate to wealth, stability, renewal, and nature.

https://www.smashingmagazine.com/2010/01/color-
theory-for-designers-part-1-the-meaning-of-color/
Cool Color: Purple

 In ancient times, the purple dyes were very expensive, so

only royals and the very wealthy could afford them.
 In design and visualization
 Dark purples can give a sense of wealth and luxury.
 Light purples are softer and are associated with spring and romance.

https://www.smashingmagazine.com/2010/01/color-
theory-for-designers-part-1-the-meaning-of-color/
Color Representation
Color Names vs Digital Representation

 Modern browsers support 147 different color names, only 16 out of which
are standard colors (W3.org).
 Standard color names are: aqua, black, blue, fuchsia, gray, green, lime,
maroon, navy, olive, purple, red, silver, teal, white, and yellow.
 147 names are derived from these names.
 Human visual system can distinguish (Gibson, et al., 2017) close to a million
different colors in nature.
 Color display or print devices can reproduce hundreds of thousands of
colors.
 Hence a digital representation when a larger number of colors or a color
set are needed.
RGB Color Model
Additive Mixing Red, Green and Blue Colors

Additive Color Mixing

 The RGB Cube

 Represents color using 3 color channels

 The range of values in each channel is [0 to 1]
 Internally in a device is converted to a unsigned byte
representation. So internal range is [0 to 255]

 A color is represented as a triplet

 (r, g, b): (0.20, 0.66, 0.32) or (50, 168, 82)
 Hex representation: #32a852
RGB Color

 Pro:
 A large number of colors can be represented
 Total colors possible: 256 x 256 x 256
 Cons:
 Not intuitive
Artist’s Color Models

 Color Mixing:
 The mixing of color “pigments”

 ex:
 Red, Yellow, Blue (Artist paint)
 Cyan, Magenta, Yellow (printing ink)
 Color Wheel
 Artists Organize Color in a Color Wheel
 Primary Colors
 Red,
 yellow, and
 blue
 Secondary Colors
 violet (or purple),
 orange, and
 green (VOG)
 formed by mixing equal amounts of red
and blue, red and yellow, and blue and
yellow, respectively
 Hues:
 Colors in Color wheel, also called Pure
colors https://drawpaintacademy.com/a-comprehensive-guide-
to-color-theory-for-artists/
Hue: Pure Colors

 Warm Colors in one half

 Cool colors in the other
Shades, Tints and Tones

 Tints:
 Add white to Pure color
 Shades:
 Add black to Pure color
 Tone:
 Add gray to Pure color

https://en.wikipedia.org/wiki/Tints_and_shades
Color Solid

 3D extension of the color wheel for color representation.

 A single color wheel, but multiple color solids!
 HSV: Hue, Saturation and Value
 HSL: Hue, Saturation and Lightness

https://en.wikipedia.org/wiki/File:Hsv_color_circle.svg
Color Representation in HSL and HSV

 Both use Cylindrical Coordinate system

 H, the angular coordinate: color in degrees and counterclockwise direction on
color wheel. Range 0 to 360 degrees.
 0(360) deg - red
 120 deg - green
 240 deg - blue
 S, the radial coordinate, is specified in percent. Range 0 to 100%
 0% at the center
 100% at the periphery
 V (for HSV) or L (for HSL), the axial coordinate: also specified in percent. Range 0
to 100%
 0 % is at the bottom, represents black.
 100% is at the top, represents white
HSL vs HSV

 HSV model:
 the V value of 100% means all colors and tones are at their brightest
form,
 HSL model:
 the L value 100% means all colors
 tints are indistinguishable from white.
 HSL model attempts to match the lightness of all the pure colors and
their tones to the same value.
HSL and HSV

HSL cylinder (on left) and HSV cylinder(on right)

https://en.wikipedia.org/wiki/HSL_and_HSV
HSL and HSV

HSL double cone(on left) and HSV single cone(on right)

https://en.wikipedia.org/wiki/HSL_and_HSV
Summarizing Color Terms
 Hue is color (blue, green, red, etc.).
 Saturation refers to how strong or weak a color is (high saturation
being strong).
 Value refers to how light or dark a color is (light having a high
value).
 Tones:
 created by adding gray to a color, making it duller than the original.
 Shades:
 created by adding black to a color, making it darker than the original.
 Tints:
 created by adding white to a color, making it lighter than the original.
(recap) Color Preference
(recap) Warm/Cool colors and
their careful use
(recap) RGB Cube: Display color
(recap) Intuitive Color Model

Color Wheel and HSL color solid

(recap) Color: Hue, Tint, Shade

shades of a single hue, ideal for

sequential data.

https://www.franklinpainting.com/blog/home/choosing-colors-interior-
painting/
Color Wheel: Analogous colors

 Colors that are adjacent on the color wheel, for example, blue and
blue-green, orange and red-orange.
Color Wheel: Complimentary Color
 Picking the right color

 Comparing two variables:

 Comparing with no value
judgment: analogous

 Highlighting opposite
property: Complementary

 when one is of primary

focus: near complementary

https://www.storytellingwithdata.com/blog/2020/5/6/picking-the-right-colors
 Picking the right color (contd..)

 Showing quantitative value:

 changes in value from zero
to a maximum value:
sequential color ramp
 Lowest matching the
background color
 Highest matching the key
color.
 changes in value through a
range with a meaningful
midpoint: diverging color
ramp
 Say, Data includes negative
and positive numbers. Ex:
correlation coefficient.

https://www.storytellingwithdata.com/blog/2020/5/6/picking-the-right-colors
Color and Human Eye
Human Visual System

From : SPIE Commercial + Scientific Sensing and Imaging, At Los Angeles, Volume: 10223
The Eye
Human Retina

The Retina : An Approachable Part of the Brain by J.E. Dowling, Belknap Press, page: 14-15
Retina

• Converts light in to signals

• Transmits signal to brain

# cones in each retina: 5 x 106

# rods in each retina: 108
# optic nerves : 1.5 x 106

Light
http://white.stanford.edu/~brian/numbers/numbers.html
Retina: Photoreceptors

• Cells containing photosensitive pigments

• Two types of photoreceptors: rods, cones
• foveal region (~2 degrees) has very high
resolution of photoreceptors (mostly cones).
• Peripheral region has low resolution of
photoreceptors (mostly rods)

Cones are most densely packed

within a region of the retina
called the fovea
Rods

Number of rods in each retina: 108

Photoreceptors which

• generate achromatic response

• Sensitive only in low light levels

(scotopic range)
Cones

Number of cones in each retina: 5 x 106

Photoreceptors

• responsible for color vision

• 3 Types of cones: S, M, L
In Fovea (64%L, 32%M, 2%S)

• Sensitive only in high light levels (photopic range)

Cones : S, M and L

• S : Blue Sensitive (peak: 445nm)

• M : Green or yellow Sensitive

(peak 535nm)

• L : Red Sensitive (peak: 575nm)

• Not uniformly sensitive
- more sensitive to green than
red
Three Color Response Signals from
Retina

For a spectrum of light I():

Response form L cone: l =  I()L()d

Response from M cone: m =  I()M()d
Response from S cone: s =  I()S()d
where
L(), M(), and S() are the sensitivity
curves of the three sets of cones.
Three Color sensors: Metamers

3 numbers?
 We can’t distinguish certain colors
Such colors are metamers of each other
Taking advantage of metamerism

P P

X X

 400 700

400 700

S1,M1 ,L1 = S2 ,M2 ,L2

Visual match!
Dogs have two cones

 Unlike humans who have three different color sensitive cone cells in
their retina (blue (S), green (M) and red (R)) dogs have only two
(blue, yelllow).
 Dogs can not distinguish yellow or red objects based on their color.

Ref: https://dog-vision.com/
Color Deficieny
Cone Photoreceptors

• S : Blue response (peak:

445nm)

• M : Green or yellow response

(peak 535nm)

• L : Red response (peak:

575nm)

• deficiency of one or more

cone type results in
colorblindness
(~ 8% caucasian males are color
deficient)
Issues

 Information in documents and presentations can be lost to people

with even a mild form of colour blindness.
 Most businesses are completely unaware of the issues which can
affect them arising from the high incidence of color blind people
(especially men, 8%) in the population.

http://www.colourblindawareness.org/business/
Color Deficiency/Blindness

Most color deficients can see a limited number of colors

Normal Vision Deuteranopia

Color Deficiency: a male problem

 In India estimated
 8% of male population
 less than 1% female population : ~0.4%
 Colorblindness is most present in males due to the way genetics
work.
 Red and green color deficiency being the most common form of
color-blindness
The Supreme Court has directed the Film and Television Institute of India (FTII) not to exclude
candidates suffering from color blindness from its courses on film making and editing and
asked it to make changes to its curriculum instead.

Source:
https://indianexpress.com/article/explained/everyday-explainers/what-is-colour-blindness-7867542/
April 13, 2022
Color Deficiency

Souce: Mayo Clinic

Color Deficiency/Blindness

 Most color deficients can see a limited number of colors

 a.k.a Daltonism: Named after John Dalton
 John Dalton famous chemist, physicist, and meteorologist.
 Best known for proposing the modern atomic theory and for his research
into color blindness.
 In 1794, Dalton wrote his paper on "Extraordinary facts relating to the
vision of colours",
 he postulated that shortage in colour perception was caused by
discoloration of the liquid medium of the eyeball.
 as both he and his brother were colour blind, he recognised that the
condition must be hereditary.
 Resources: http://www.colourblindawareness.org
Types of Color Deficiency

 Red-Green Color Blindness: Most common. Involves the loss

or limited functions of the red cone photo pigments (also
called protan) or green cone pigment (also called
deutran).
 Protanopia: Complete Loss of Red cone. Certain shades of
orange, yellow, and green all appears as yellow. (1% of males)
 Protanomaly: Red cone is weak. Red, orange, and yellow all
appears greener and not as bright. (1% of males, 0.01% of
females)

Wiki
Normal Vision

www.color-blindness.com/coblis1-color-blindness-simulator/
Protanopia

Red-Blind Red-Weak

www.color-blindness.com/coblis1-color-blindness-simulator/
Types of Color Deficiency

 Red-Green Color Blindness: Most common. Involves the loss or

limited functions of the red cone photo pigments (also called
protan) or green cone pigment (also called deutran).
 Protanopia: Complete Loss of Red cone. Certain shades of
orange, yellow, and green all appears as yellow. (1% of males)
 Protanomaly: Red cone is weak. Red, orange, and yellow all appears
greener and not as bright. (1% of males, 0.01% of females)
 Deuteranopia: Loss of Green cone. Red appears as brownish
yellow and green appears as beige. (1% of males. Actual
Daltonism)
 Deuteranomaly: Green cone is weak, meaning yellow and green
appears redder and it is difficult to tell violet from blue.
 Most common of Red-Green color blindness. 6% of males, 0.4% of
females

 Note: Protanomaly leads to loss of brightness whereas

Deuteranomaly does not. Wiki
Deuteranopia

Green-Blind Green-Weak

www.color-blindness.com/coblis1-color-blindness-simulator/
Types of Color Deficiency
(continued)
 Blue/yellow Color Blindness: Rarer than red/green color blindness. It
involves the missing or limited functions of the blue cone photo
pigment (also called tritan).
 Tritanopia: Loss in Blue cone. Blue appears greener and yellow appears
violet or light grey. (<1% of males and females)
 Tritanomaly: Blue cone is weak: Blue appears greener and it is difficult to see
yellow and red from pink.

 Not gender specific.

Tritanopia

Blue-Blind Blue-Weak

www.color-blindness.com/coblis1-color-blindness-simulator/
Types of Color Deficiency
(continued)
 Monochromacy: Complete inability to see color.
 only a single channel for conveying information about color.
 Two types:
 Rod monochromacy: no cones
 vision in lights of normal intensity is difficult. While normally rare
 Cone monochromacy: One type of cone
 good pattern vision at normal daylight levels, but can not distinguish hues
 Rare among humans
 Many animal species, such as marine mammals, the owl
monkey and the Australian sea lion are monochromats.
Diagnosis: Ishihara color test.

The test most often

used to diagnose red–
green color
deficiencies

Wiki
Different Color Visions

Protanopia- Red cone is missing Deuteranopia- Green cone is missing

Protanomaly- Red cone is weak Deuteranomaly- Green cone is weak
Tritanopia- Blue cone is missing
Tritanomaly- Blue cone is weak
Different color visions

https://wearecolorblind.com/articles/a-quick-introduction-to-color-blindness/
(recap) Color Deficiency

 In India estimated
 8% of male population
 less than 1% female population : ~0.4%
 Colorblindness is most present in males due to the way genetics
work.
 Red and green color deficiency being the most common form of
color-blindness

Source:
https://indianexpress.com/article/explained/everyday-explainers/what-is-colour-blindness-7867542/
April 13, 2022
(recap) Different color visions

https://wearecolorblind.com/articles/a-quick-introduction-to-color-blindness/
Qualitative color scale:
Color for Categorical data
 Qualitative color scale: a finite set of specific colors that are chosen
 To look clearly distinct from each other while also being equivalent to
each other
 no one color should stand out relative to the others
 the colors should not create the impression of an order, as would be the
case with a sequence of colors that get successively lighter.
Plotly: Qualitative Color choices
Plotly: Qualitative Color choices

px.strip(df, y="total_bill", x="day", color="day",

color_discrete_sequence= px.colors.qualitative.Dark2)
Seaborn: Qualitative color choices
Seaborn: Using different color scale

sns.catplot(x="day", y="total_bill", data=tips, hue = "day",

palette=sns.color_palette("dark"))
Sequential Color Scale:
Color for Quantitative data
 a sequence of colors that clearly indicate
 which values are larger or smaller than which other ones and
 how distant two specific values are from each other.
 the color scale needs to be perceived to vary uniformly across its entire
range.
 Sequence is mostly darker at one end lighter at the other end.
Sequential Color Scale:
Color for Quantitative data
Sequential Color Scale:
Color for Quantitative data

Seaborn Sequential Color palettes

Sequential Color Scale

Source: https://www.mapsofindia.com/maps/india/population.htm
Diverging Color Scale:
Color for Quantitative data
 Two sequential palettes stitched together at a common midpoint,
which usually is represented by a light color

 The perceived magnitude of a data value would depend on whether it

fell above or below the midpoint value.
Diverging Color Scale

px.choropleth(…
color_continuous_scale=px.colors.diverging.BrBG,
color_continuous_midpoint=avg_lifeExp,
title=…)
How to Choose Color?

 Color Brewer: http://colorbrewer2.org

 Chroma.js : https://github.com/gka/chroma.js/
 Adobe Color CC: https://color.adobe.com/create/color-wheel/
 I want hue: http://tools.medialab.sciences-po.fr/iwanthue/
Color Brewer

 An Online tool for choosing effective schemes for thematic maps

 Author: Cynthia A. Brewer
 an American professor of geography at the Pennsylvania State
University, and author.
 Widely used by most visualization tools:
 ggplot, MatplotLib, plotly, D3…

http://colorbrewer2.org
Colorbrewer
Colorbrewer: How to use?

 Nature of the data:

 Sequential: Increasing order or decreaing order.
 Divergent: Negative and positive
 Qualitative: Nominat/Categorical data (no order)
Colorbrewer: How to use?

 Sequential Scheme: used to  Human Psychology associates:

represent  darker color to higher density
 quantitative: ex: Unemployment,  higher density with greater
obesity rate. numbers.
 ordinal: ex: least polluted to  So, dark colors are perceived as
most polluted; least desirable being higher in value than lighter
vacation spot to most desirable ones.
vacation spot.
 Mapping Strategy:
 map the large end of you scale to
a dark color and the smaller end
to a light color.
 The bigger the difference between
these two extremes, the more
effective your use of color.
Colorbrewer: How to use?
Sequential Scheme: Color is generated by combining value with Hue
Darker color for larger value, lighter for smaller value

Single Hue

https://theblog.adobe.com/the-power-of-the-palette-
why-color-is-key-in-data-visualization-and-how-to-use-it/
Colorbrewer: How to use?
Sequential Scheme: Color is generated by combining value with Hue
Darker color for larger value, lighter for smaller value

Multiple Hue
Colorbrewer: How to use?

 Diverging Scheme:
 used to represent sequence, but in this
case there is a meaningful mid-point to
our scale.
 useful for making comparisons with
some critical value in the data. Order
 Ex: Positive-Negative population growth
compared to certain base line (could
be mean)
 Ex: greenhouse gas emission reductions
to emphasize how much more some
countries have reduced their emissions
beyond the target specified in a treaty
 can be constructed using two sequential
schemes fused at the middle
 Can have a larger number of levels
compared to Sequential scheme

https://theblog.adobe.com/the-power-of-the-palette-
why-color-is-key-in-data-visualization-and-how-to-use-it/
Colorbrewer: How to use?

Diverging Scheme
Colorbrewer: How to use?

 Qualitative Scheme: Useful for

categorical data (different type of
phenomenal, different countries)
 Ex: slices on a pie chart, multi-line
chart, bar chart.
 Adajacent hues could be a good
choice.
 Disadvantage: When viewed on a
projector, or seen by visually
impaired, colors look very similar.

https://theblog.adobe.com/the-power-of-the-palette-
why-color-is-key-in-data-visualization-and-how-to-use-it/
Colorbrewer: How to use?

 Qualitative Scheme: Useful for

categorical data (different type of
phenomenal, different countries)
 Ex: slices on a pie chart, multi-line
chart, bar chart.
 Better chioce is: Choose adjacent
hues, but take larger steps (at least
35° of hue change) and maintain
a consistent lightness.
 any item should not stand out
because of the lightness difference

https://theblog.adobe.com/the-power-of-the-palette-
why-color-is-key-in-data-visualization-and-how-to-use-it/
Colorbrewer: How to use?
Qualitative Scheme
Colorbrewer
Common pitfalls of color use
encoding too many different items in different colors

Bad
 Common pitfalls of color use

Better

qualitative color scales work best when there are only a

few < 6 different categories to be colored.
Common pitfalls of color use
coloring for the sake of coloring

Bad
Common pitfalls of color use

Good
Common pitfalls of color use

 Not using color-blind safe colors.