Colour temperature is usually the first quantitative concept encountered when studying lighting, yet it is also one of the most misunderstood. While it provides a useful shorthand for describing the overall chromatic bias of a light source, it says almost nothing about how that light will interact with real objects, surfaces, and materials. Understanding why colour temperature is insufficient requires a deeper examination of how light is produced, how colour is revealed, and how measurement systems evolved to address the limitations of simple descriptors.
Colour temperature is derived from the concept of a black-body radiator, a theoretical object that emits electromagnetic radiation when heated, producing a predictable spectrum at each temperature. As temperature increases, the emitted light shifts from red through orange and yellow toward white and blue. This model provides a convenient reference because it links colour to a physical process rather than subjective appearance. Tungsten filament lamps approximate black-body behaviour reasonably well, which is why traditional incandescent lighting aligns closely with its rated colour temperature.
However, most modern light sources do not behave like black-body radiators. Fluorescent lamps, discharge lamps, and especially LEDs produce light through electronic excitation rather than thermal emission. Their spectral output is therefore discontinuous, composed of peaks generated by phosphors or semiconductor junctions rather than a smooth curve. Two light sources may share the same correlated colour temperature while having radically different spectral power distributions. This discrepancy is the root of many colour rendering problems.
Colour temperature describes the colour of the light itself, not its ability to reveal the colours of objects. This distinction is critical. A light source may appear neutral white when measured on a Kelvin scale yet render skin tones poorly, distort fabrics, or suppress subtle colour differences. White balance adjustments can compensate for overall colour bias, but they cannot reconstruct missing spectral information. Once a wavelength is absent from the source, it cannot be reflected by a surface, and therefore cannot be recorded.
The need to evaluate how light reveals colour rather than merely how it appears led to the development of colour rendering metrics. The most widely adopted of these is the Colour Rendering Index, or CRI.
CRI was developed by the International Commission on Illumination (CIE) as a standardised method for comparing the colour rendering performance of a test light source against a reference illuminant. The reference is chosen based on colour temperature: daylight for higher temperatures and tungsten-like sources for lower ones. The fundamental question CRI seeks to answer is whether objects appear the same under the test light as they would under the reference.
The CRI calculation involves illuminating a set of standardised colour samples and measuring how their appearance shifts under the test source relative to the reference. These samples, known as the Ra set, consist of eight low-saturation pastel colours designated R1 through R8. The choice of pastel tones was deliberate, as they provide stable, repeatable measurement conditions and minimise extreme spectral interactions. The differences between the test and reference renderings are quantified mathematically, and the resulting score is expressed on a scale from 0 to 100, where 100 represents perfect agreement with the reference.
A CRI score of 100 indicates that the test source renders the chosen samples identically to the reference illuminant. Scores in the high 90s are considered excellent, while scores below 80 typically indicate noticeable colour distortion. Importantly, CRI is not a linear perceptual scale. A drop from 95 to 90 is not equivalent to a drop from 85 to 80 in visual terms, and different colours may be affected unevenly.
Measurement of CRI requires spectrophotometric analysis of the light source. Institutions such as the National Institute of Standards and Technology (NIST) maintain calibration standards to ensure consistency in measurement equipment and procedures. Lighting designers and engineers often refer to recommended practices published by the Illuminating Engineering Society (IES) when specifying acceptable CRI levels for different environments.
Despite its widespread use, CRI has significant limitations, particularly for modern lighting technologies. One of the most frequently cited criticisms is the restricted nature of the Ra colour samples. The eight pastel tones do not adequately represent the saturated colours commonly encountered in real scenes, such as deep reds, blues, and complex skin tones. As a result, a light source may score highly on CRI while performing poorly in practical applications.
To address this, extended CRI evaluations sometimes include additional test colours, such as R9, which represents a strong red sample. R9 performance is particularly important for rendering skin tones accurately, yet it is not included in the standard Ra average. Many LED lights advertise high CRI values while exhibiting poor R9 performance, leading to flattering specification sheets but disappointing real-world results.
Research institutions, including Rensselaer Polytechnic Institute, have demonstrated that CRI’s reliance on limited samples and human-vision-based weighting makes it an imperfect predictor of performance for imaging systems. Cameras do not respond to colour in the same way the human eye does, and the spectral sensitivities of sensors differ from photopic vision curves. As digital imaging became dominant, the need for camera-specific evaluation metrics became increasingly clear.
This need led to the development of the Television Lighting Consistency Index (TLCI), introduced by the European Broadcasting Union. TLCI was designed specifically to assess how well a light source performs when captured by television cameras rather than evaluated by human observers. Instead of comparing colour appearance perceptually, TLCI models camera sensor responses and calculates how much colour correction would be required to achieve a reference appearance.
TLCI scores range from 0 to 100, but their interpretation differs from CRI. A high TLCI score indicates that little to no colour correction would be required in practice, while lower scores suggest increasing difficulty in achieving acceptable colour reproduction. This shift from perceptual accuracy to correction effort represents a fundamental change in evaluation philosophy. Rather than asking whether colours look the same, TLCI asks how problematic a light source will be in use.
TLCI also incorporates a broader set of colour samples, including more saturated hues, and evaluates performance across different camera profiles. This makes it a more practical tool for modern production environments, particularly those relying on digital sensors with known response characteristics.
More recently, spectral similarity metrics such as SSI have emerged, further refining the evaluation of light sources. SSI compares the spectral power distribution of a test source directly to that of a reference, producing a numerical measure of similarity. Unlike CRI, which compresses complex spectral behaviour into a single averaged score, SSI preserves information about how closely two spectra align across the entire visible range.
Spectral metrics acknowledge a critical reality: colour rendering is fundamentally about spectral completeness and balance. A light source with a smooth, continuous spectrum will generally perform better across a wide range of materials than one with narrow peaks and gaps, even if both share the same colour temperature. Spectral analysis therefore provides the most physically grounded assessment of lighting quality.
Understanding these metrics requires recognising that no single number can fully describe a light source. Colour temperature, CRI, TLCI, and SSI each address different aspects of performance, and each has contexts in which it is useful or misleading. Advanced lighting practice involves selecting and interpreting these metrics appropriately rather than relying on marketing specifications.
It is also important to understand the economic and technological forces shaping these standards. Manufacturers often optimise products to perform well on specific metrics, sometimes at the expense of broader performance. High CRI scores can be achieved through spectral tuning that favours the Ra samples while neglecting other wavelengths. Awareness of this dynamic allows practitioners to read specifications critically rather than accepting them at face value.
In practical terms, the evaluation of lighting quality increasingly relies on a combination of published metrics, independent testing, and direct observation using reference materials. Colour charts, skin tone evaluation, and comparative testing remain essential tools alongside numerical indices. Metrics guide decision-making, but they do not replace informed judgement.
The progression from colour temperature to CRI, then to TLCI and spectral metrics, reflects a broader shift in lighting practice: from descriptive simplicity toward analytical precision. As lighting technologies evolve, evaluation methods must evolve alongside them. Advanced practitioners are those who understand not only how to read these metrics, but why they exist, what assumptions they make, and where they fail.
Only with this understanding can lighting decisions be made deliberately rather than reactively, grounded in physical reality rather than marketing language.