Colour and memory

What it is

An experiment on whether colour makes everyday objects easier to remember. The work started as an assignment in the course IMT4315 Colour in interface design at NTNU in spring 2025. Supervisor Phil Green selected the paper for AIC 2025, we revised it together, and he gave the presentation in Taipei.

AIC, the International Colour Association, brings together the national colour science societies. AIC 2025 was the organisation's 16th congress, hosted by the Color Association of Taiwan.

The brief

Prior research on colour and memory is inconsistent. Some studies find better recall with colour-coded information, others find no effect. Much of the inconsistency may come down to experimental design: stimuli with an internal relationship let participants guess forgotten objects from the ones they remember, and the time in each phase is often poorly controlled. The question became: does colour help recall once the chance to guess from context is removed?

Try the test

What I did

The test design built on Kim's Game, a classic memory exercise. An earlier variant used two familiar scenes, a market stall and an office. I piloted that variant first. Participants reasoned their way to forgotten objects from the ones they remembered, because the objects in each scene belonged together. The scenes were therefore replaced with AI-generated images of everyday objects with no relationship to each other, one in colour and one in greyscale.

Twenty young adults took part, aged 20 to 28. All of them completed the Ishihara test for colour vision.

I ran every round myself, with the same sequence each time. Fifteen seconds of exposure to the image. Ten seconds of counting backward, to block internal rehearsal. Sixty seconds of spoken recall, recorded on audio. Only exact matches counted as correct, and the order of colour and greyscale was randomised between participants.

For the colour image I measured four properties per object: lightness, chroma, relative size and entropy. RGB values were converted to CIELCh, and each property was correlated against the proportion who remembered the object.

The result went against expectation. Participants remembered 45 percent of the objects on average, the same for colour and greyscale. None of the colour properties correlated with recall. Relative size did, with a correlation of 0.53. For these images, space mattered more than colour.

What I would change

The same objects in both conditions. The two images contained different objects, and the greyscale image had 17 unique ones against the colour image's 20. That means the measurement isn't just of colour, but of the objects too. The next version would use the same image in a colour and a greyscale version.

More participants and more rounds. With 20 participants and one image per condition, the sample is too small to rule out small effects. The conclusion holds for this study, not for the question as a whole.

Control for placement. Western participants read an image in a Z pattern, and objects in the top left have an advantage. Placement should have been balanced or measured as its own variable.

Facts

Role

Design, pilot testing, execution and first draft, Mathias Sætersdal. Supervision and revision, Phil Green, co-author.

Course

IMT4315 Colour in interface design, NTNU, spring 2025. Grade A.

Selected

The paper was one of two from the course selected for AIC 2025.

Presented

At AIC 2025, the 16th congress of the International Colour Association, Taipei, 19 to 24 October 2025. Phil Green gave the presentation. I did not travel.