The Scientific Foundation of the Rainbow Controversy
To understand the current debate, one must first understand the technical grievances held against rainbow colormaps. In a standard rainbow scale, data values are mapped to the hues of the visible light spectrum: violet, indigo, blue, green, yellow, orange, and red. The primary scientific objection is the lack of "perceptual uniformity." In an ideal colormap, equal changes in data values should correspond to equal perceived changes in the visualization. Rainbow scales fail this test because human eyes are more sensitive to certain wavelengths than others. For example, the transition from yellow to green appears much more dramatic than a similar numerical shift within the blue range.
Furthermore, the luminance, or perceived brightness, of a rainbow colormap is non-monotonic. It fluctuates wildly; yellow is perceived as significantly brighter than the blue or red that may flank it. This creates "false edges" or "artifacts" in the data—visual boundaries that do not exist in the underlying numbers but appear to the viewer because of the sudden change in brightness. Despite these well-documented flaws, the rainbow remains the default choice for high-stakes visualizations, ranging from National Oceanic and Atmospheric Administration (NOAA) weather maps to medical imaging and fluid dynamics simulations.
A Chronology of the Anti-Rainbow Movement
The academic campaign against the rainbow colormap has spanned nearly four decades, evolving from niche technical observations to a standard principle of design.
- The Early 1990s: As computational power grew, software like MATLAB began using the "Jet" colormap as a default. Its high contrast made it popular for displaying complex datasets, even as perceptual psychologists began warning about the distortion of data.
- 2007 – The Turning Point: David Borland and Russell M. Taylor II published "Rainbow Color Map (Still) Considered Harmful." This paper became the manifesto for the anti-rainbow movement, detailing how the colormap obscured features and misled viewers.
- 2014-2015 – The Rise of Viridis: The Python programming community, specifically the Matplotlib development team, sought to replace Jet. They engineered "Viridis," a colormap that is perceptually uniform, monotonic in luminance, and accessible to colorblind users. This sparked a global shift toward "scientific" colormaps in academic publishing.
- 2023 – The Reassessment: The paper "Rainbow Colormaps Are Not All Bad" marks a significant shift in the timeline. Authors Ware, Stone, and Szafir—representing both the foundational and modern eras of color research—called for a more nuanced approach, suggesting that the "finger-wagging" had ignored practical utility.
Supporting Data: When the Rainbow Outperforms the Alternatives
The core of the 2023 research lies in task-based analysis. The authors argue that while rainbow colormaps are poor for perceiving the "shape" of data (the 3D-like contours of a surface), they are exceptionally effective for "value reading."
In a value-reading task, a user looks at a specific point on a map and attempts to determine its numerical value by comparing it to a legend. Because a rainbow colormap uses a wide variety of distinct hues, the human brain can more easily categorize and recall specific points on the scale. For instance, on a NOAA temperature map, "bright yellow" might represent exactly 90 degrees Fahrenheit. Because yellow is visually distinct from the surrounding orange and green, the user can quickly identify the 90-degree threshold.
In contrast, perceptually uniform maps like Viridis or Magma often use a single or dual-hue gradient (e.g., dark purple to bright yellow). While these are mathematically superior for showing smooth transitions, they make it harder for a viewer to distinguish between, for example, the 60% and 65% marks on a scale, as the colors are too similar to the naked eye. Research cited by the authors suggests that for tasks requiring the identification of specific quantities, the rainbow scale consistently yields higher accuracy and faster response times among test subjects.
Analyzing the "Shape Perception" Strawman
One of the most frequent criticisms of the rainbow is that it fails to convey "shape." When a rainbow scale is applied to a complex 3D surface, the fluctuations in luminance make it impossible for the human visual system to interpret the shading and contours of the object. To illustrate this, researchers often apply a rainbow colormap to a 3D model of an apple or a human face, resulting in a distorted, unrecognizable image.
However, Ware, Stone, and Szafir argue that this is often a "strawman" argument. In many practical applications of data visualization, such as a 2D map of sea-surface temperatures or a flat atmospheric pressure chart, the "shape" of the data is not the primary concern. The goal is often to identify "global structure"—where the hot spots are, where the cold fronts are moving, and the relationship between different distributions. The high contrast of the rainbow draws attention to these global patterns more effectively than the subtle gradients of "safer" colormaps. By acknowledging that shape perception is not always the goal, the authors suggest that the rainbow’s flaws are, in many contexts, irrelevant.
Official Responses and Practical Implications
The visualization community has reacted to these findings with a mix of caution and validation. Practitioners in fields like meteorology and oceanography, who have long resisted the academic push to abandon rainbows, find support in the paper’s conclusions.
Meteorologists at NOAA, for example, have noted that public-facing maps must be instantly readable by a general audience. The rainbow scale utilizes common cultural associations (blue for cold, red for hot) that are deeply embedded in the public consciousness. Replacing these with a purple-to-yellow Viridis scale could lead to confusion during critical weather events. The paper suggests that rather than forcing a change in colormap, research should focus on "optimizing" the rainbow—adjusting the luminance of the green and yellow segments to minimize artifacts while retaining the high-contrast hue transitions.
From a software development perspective, the implications are significant. Major data visualization libraries are now considering "multi-hue" scales that bridge the gap between the traditional rainbow and the perceptually uniform models. These "improved rainbows" aim to maintain the high-resolution value reading of the spectrum while smoothing out the luminance spikes that cause visual distortion.
Broader Impact on Scientific Communication
The debate over colormaps extends beyond aesthetics; it impacts how scientific data is interpreted by policymakers and the public. In medical diagnostics, for instance, the use of a specific colormap in an MRI or PET scan can influence a clinician’s ability to detect a tumor. If a colormap is too smooth, a subtle but critical boundary might be missed. If it is too high-contrast (like a poor rainbow), a false boundary might be diagnosed as a problem.
The work of Ware, Stone, and Szafir encourages a move away from "one-size-fits-all" design rules. It suggests that the choice of a colormap must be "task-dependent."
- For Shape Perception: Use perceptually uniform, monotonic luminance scales (e.g., Viridis, grayscale).
- For Value Reading: Use multi-hue scales or optimized rainbows that provide distinct markers for the eye.
- For Global Structure: Use high-contrast scales that highlight extremes and distributions.
Conclusion: Finding the Middle Ground
The paper "Rainbow Colormaps Are Not All Bad" serves as a call for a more sophisticated understanding of human perception in data science. By moving past the era of "finger-wagging," the visualization community can begin to treat the rainbow not as a mistake to be erased, but as a tool to be refined. The authors conclude with a mandate for future research: "We understand well enough why rainbows can be bad; let us focus instead on finding out when and why they are good."
As data becomes increasingly central to public life, the ability to visualize it accurately and effectively is paramount. The defense of the rainbow colormap is not a defense of bad design, but a recognition that human cognition is complex. Sometimes, the "derided" tool is the most effective one for the job, provided the designer understands the strengths and weaknesses of the medium. The future of data visualization likely lies in a hybrid approach—one that respects the mathematical rigor of perceptual uniformity while embracing the communicative power of the full visible spectrum.