PSLE Science Reality Lab Vol No.214 | “CRI = 90” — Does That Mean Colours Are 90% Accurate?

PSLE-SCI-REALITY-0214

Wait, What? Two Lamps Both Say “White” — So Why Does the Red Apple Change?

Imagine a lighting shelf with two LED bulbs. Both are labelled 3000 K. Both produce about the same amount of light. One box also says CRI 90. Under the first lamp, a red apple looks rich and familiar. Under the second, the same apple looks duller. A learner reads “CRI 90” and says, “Easy. The colours are 90% accurate.”

That answer sounds reasonable because 90 looks like a percentage. But the label does not say 90%. It names a colour-rendering metric. Traditional general Colour Rendering Index, often written CRI or Ra, compares how selected test colours appear under the test light with how they appear under a suitable reference illuminant. A higher score generally means closer colour fidelity for the test samples used by the metric. It is not a statement that every colour is 90% correct, that 90% of objects look right, that the lamp is 90% bright, or that 90% of its electrical energy becomes useful light.

Reality Lab habit: when a familiar-looking number sits beside an unfamiliar metric, identify the measurement job before turning the number into a percentage story.

Quick Answer

CRI 90 is not “90% colour accuracy”.
CRI is a method-based score about how a light source renders selected object colours compared with a reference illuminant.
General CRI compresses several colour comparisons into one summary number, so the same overall score can hide differences in how particular colours are rendered.
CRI is not brightness. Lumens answer a different question.
CRI is not colour temperature. Kelvin on a lamp package answers a different question about the apparent colour of the white light.
CRI is not electrical efficiency, energy use, lifetime or preference.
A high CRI supports a bounded claim about colour fidelity under the metric; it does not prove every object will look identical to daylight or that every observer will prefer the light.

The Exact Learner Job This Volume Owns

This volume owns one real-world evidence-transfer job: how to evaluate a lamp box, product comparison, showroom card or lighting infographic that reports a CRI value without turning the score into a percentage of colour accuracy, brightness, preference or efficiency.

It does not become a full lesson on human vision, spectra, LEDs, colour spaces or lighting engineering. Those scientific concepts belong to their existing owners. Reality Lab is interested in the communication object: a single score that looks simpler than the measurement system behind it.

Rebuild the Label as an Evidence Object

Consider an original fictional lamp package:

Package line	What it is trying to tell you	What it does not automatically tell you
800 lm	Total luminous flux emitted by the lamp	Illuminance on every surface
3000 K	Correlated colour temperature of the white light	Physical operating temperature of the LED
CRI 90	Colour-rendering score under the stated metric	90% colour accuracy for every object
9 W	Electrical power rating	Colour fidelity

All four numbers can be correct at the same time because each has a different scientific job. The first discipline is therefore quantity identity: do not allow one number to answer a question assigned to another number.

Observed, Measured, Calculated, Claimed and Inferred

Observed: coloured samples can look different under different light sources.
Measured: a lighting measurement system characterises the test source and its spectral output.
Calculated: the CRI method compares colour appearance for specified test samples under the test and reference illuminants and combines the differences into a score.
Supported claim: under this metric, the source has a general colour-rendering score of 90.
Reasonable inference: the source is expected to render the metric’s test colours with relatively high average fidelity compared with the reference.
Unsupported leap: every colour is 90% accurate.
Unsupported leap: 10% of colours will be wrong.
Unsupported leap: people will prefer the lamp 90% of the time.

Why “90” Is Not Automatically “90%”

Numbers do not carry their own meaning. A score can be built from comparisons, weighting rules, reference conditions and calculations. Turning every score into “percent correct” silently changes the denominator. For a true percentage claim, you need to know what the whole is and what fraction of that whole is being counted or measured.

With CRI, the learner should ask a different set of questions: What colour samples are being compared? What reference illuminant is used? How are differences summarised? Is the number general CRI, a special colour sample score, or a different modern colour-rendering metric?

A Summary Score Can Hide Different Colour Behaviours

Imagine a made-up teaching example with eight colour-test errors. Lamp A has small errors across all eight. Lamp B has extremely small errors on seven samples but a much larger error on one red sample. If the method compresses several comparisons into one overall summary, the final scores could still look similar even though the visual weakness is distributed differently.

This is why one-number summaries are useful but incomplete. A summary is not dishonest merely because it hides detail; that is what summaries do. The scientific job is to know when the hidden detail matters to the claim you want to make.

Worked Case 1: “CRI 90 Means 90% of Colours Are Correct”

Tempting reasoning: 90 looks like a percentage, so 90 out of every 100 colours must appear correctly.

Repair: CRI is a score derived from colour differences for specified test samples relative to a reference. The number is not a count of correct colours out of 100.

Worked Case 2: “CRI 95 Must Be Brighter Than CRI 80”

Repair: brightness-related quantities and colour-rendering fidelity are different jobs. A lamp can have a high CRI and low lumen output, or a lower CRI and high lumen output. Compare lumens when you want total light output and CRI when you want the stated colour-rendering metric.

Worked Case 3: “CRI 90 and 3000 K Are Two Ways of Saying the Same Thing”

Repair: no. Correlated colour temperature describes the apparent tint of the white light along a warm-to-cool dimension. CRI concerns the rendering of object colours relative to a reference. Two lamps can have the same CCT and different colour-rendering behaviour.

Worked Case 4: “Higher CRI Means Everyone Will Prefer It”

Repair: colour fidelity and colour preference are not identical. NIST researchers have specifically studied limitations of CRI and the difference between fidelity and visual preference. A metric can be useful for one job without capturing every dimension of colour quality.

Worked Case 5: “Both Lamps Are CRI 90, So Their Spectra Must Be the Same”

Repair: a shared summary score does not prove identical underlying spectral distributions. Different light-source spectra can produce similar summary colour-rendering scores. The score is evidence about the metric result, not a fingerprint that uniquely identifies the source.

Worked Case 6: “The Photograph Proves Lamp A Has Better CRI”

A side-by-side social-media photograph shows fruit under two lamps. One side looks more colourful.

Repair: the photograph may be affected by camera white balance, exposure, image processing, display settings and editing. A photograph can illustrate an appearance difference, but it is not automatically a controlled CRI measurement. Ask whether the metric was measured using the appropriate method and whether the image itself was produced under matched conditions.

Representation Check: The Green Badge Problem

Suppose a package uses a large green badge reading “CRI 90+” beside a smaller line reading “3000 K, 800 lm, 9 W”. The design may make CRI look like an all-purpose quality grade. But visual prominence is not scientific scope. A large badge does not expand what the metric measures.

When reading any technical badge, separate three layers:

The printed value: what exactly is stated?
The measurement definition: what procedure or comparison created the value?
The marketing interpretation: what broader story is the design inviting you to believe?

Baseline Check: Compared With What Reference?

Colour rendition is inherently comparative. The test light is judged relative to a reference illuminant appropriate to the method. Without the reference, “accurate colour” becomes vague. Accurate compared with daylight? A heated-body reference? Another LED? The metric specifies a comparison framework so different measurements can be interpreted consistently.

This is a general scientific lesson: a comparison number is only as meaningful as the reference against which it was defined.

Method Check: Which Colour-Rendering Metric?

Modern lighting science uses more than one way to describe colour rendition. Traditional CRI remains familiar, but scientists and lighting organisations have developed additional metrics because one score cannot capture every aspect of colour fidelity, gamut and preference. NIST has published work explaining shortcomings of CRI, especially for some LED spectra, and has investigated alternative approaches.

For a Primary 5/6 learner, the important transfer is not to memorise the advanced metrics. It is to ask: what measurement system produced this number, and is the claim staying inside that system’s job?

What Evidence Would Strengthen a “Good Colour Fidelity” Claim?

a clearly identified colour-rendering metric;
measurement by a suitable method rather than a visual guess;
the relevant reference conditions;
additional colour-rendering information when a specific colour, such as deep red, matters;
repeatable measurements from an appropriate laboratory or instrument;
consistent performance across units if the claim is about a product line rather than one sample.

What Would Weaken an Over-Broad “Perfect Colours” Claim?

only one summary CRI number is supplied;
the advertisement converts the score into “percent accurate” without defining a percentage;
a photograph replaces measurement evidence;
brightness, CCT, efficiency and CRI are mixed together;
the claim is about a specific colour but only a general average score is shown;
the metric is unnamed or the test conditions are missing;
the conclusion changes from colour fidelity to personal preference without evidence.

How Far Can the Conclusion Travel?

If a trustworthy specification states that a lamp has general CRI 90, a careful conclusion is: the lamp achieved a general colour-rendering score of 90 under the stated CRI method, indicating relatively high average fidelity of the method’s test colours compared with its reference.

That does not establish that every colour is 90% correct, that the light will be preferred by every person, that every red object will be rendered equally well, that the lamp is brighter, or that it is more energy-efficient.

PSLE-Style Transfer Case: A “90/100” Plant-Colour Scanner

A fictional plant scanner produces a “leaf colour fidelity score” by comparing eight standard colour patches with reference values and combining the differences into one score. Scanner A scores 90. A student says, “It will measure the colour of 90 out of every 100 leaves correctly.”

Transfer answer: the score was produced from the eight test patches and a calculation rule. It is not evidence that 90% of future leaves will be correct. To make that claim, the scanner would need to be tested on an appropriate set of leaves with a clearly defined rule for what counts as correct.

Changed-Object Transfer: A Camera “Colour Score”

A camera review gives a phone a colour score of 92. Before saying “the phone reproduces 92% of colours accurately”, ask how the review created the score. Was it based on test charts? Human ratings? Several lighting conditions? A weighted formula? The same Reality Lab habit survives even though the object changed from a lamp to a camera.

Delayed Independent Return: Four Questions

What exactly does the number measure?
What reference or baseline is built into the measurement?
What details were compressed into the summary score?
Which tempting claims go beyond that measurement job?

Explained Practice

1. A bulb says CRI 90 and 800 lm. Which number addresses total light output? The lumen value. CRI has a different job.

2. Two bulbs both say 3000 K. Must they have the same CRI? No. CCT and colour-rendering performance are different quantities.

3. Does CRI 90 mean 10% of colours are wrong? No. The score is not a count of correct versus incorrect colours.

4. Can one CRI number hide a weakness in a particular colour? Yes. A summary can hide variation among individual colour comparisons.

5. If a product photo looks vivid, has CRI been measured? Not necessarily. Camera processing and display settings can affect the image.

6. What is the strongest safe claim from a verified CRI 90 result? That the source achieved a CRI score of 90 under that method; broader claims need additional evidence.

Parent and Tutor Teaching Guide: Separate the Four Numbers

Put four cards on a table: 800 lm, 3000 K, CRI 90 and 9 W. Ask the learner to match each to one question: “How much light leaves the lamp?”, “What tint does the white light have?”, “How faithfully are test colours rendered relative to a reference?” and “What electrical power is rated?”

Then deliberately swap the questions. Ask whether CRI can answer brightness, whether kelvin can answer electrical use, and whether watts can answer colour fidelity. The learner should reject the mismatches before doing any calculation. This trains quantity identity rather than memorisation.

Finally, show two invented lamps with equal CRI but different scores for one particular colour sample. Ask what the average concealed. The aim is not to teach advanced colourimetry; it is to teach that a summary number can be valid and still incomplete.

Why This Belongs in PSLE Science Reasoning

The 2026 PSLE Science assessment objectives include interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning. The 2023 Primary Science syllabus also advocates healthy scepticism: questioning observations, methods, processes and data, and reviewing one’s own ideas. A technical product label is exactly the kind of compact scientific communication that rewards those habits.

Authoritative Sources

The Quiet Return

The number 90 did not deceive us. We deceived ourselves when we silently added a percent sign and a denominator that were never there.

Read the metric before you read the magnitude.

Small Group Tutorials