Why is deltaE not used as a measure of error?

Question

I've recently read a few articles pertaining to visual memory. Most of them (e.g. Bays et al., 2009 use the continuous color recall task where they ask participants to recall the color of an object by clicking on a color wheel. They use as measure of error the angular distance between the participant's response and the actual angle on the color wheel of the object's color.

Why use this dependent variable? In the industry, people use deltaE to determine the color difference between a product and a prototype.

Wouldn't it make more sense to also use deltaE as dependent variable in psychology? From what I've read, deltaE tries to be accurate for human color perception.

Answer 1

Short answer
To me, it seems perfectly feasible to use deltaE both as dependent and independent variable of color difference.

Background
DeltaE (Fig. 1) can be used to generate the target stimuli by pre-selecting a set of different color pairs to use in the recall task (method-of-constant stimuli, MOC), or alternatively, by using different step sizes of deltaE along the color matching test (in case of an adaptive staircase procedure for instance). It will depend on the study setup what specific method one wishes to use (MOC or an adaptive test).

On the other hand, deltaE can very well be used during data analysis to determine the difference between the presented color and the subject's response to that color. The advantage of using a single number (deltaE) over, e.g., RGB values is that deltaE is a single number (i.e. an angle) and can hence be straightforwardly used as an outcome measure, whereas RGB values would have to be recoded along three axes and somehow transformed into a single number representing response error.

^{Fig. 1. Color wheel. source: Nguyen et al. (2018)}

<sub>Reference
- Nguyen et al., BioResources (2018); 13(3), 6250-64</sub>

Answer 2

Delta E is used extensively in vision research, however the study you sited only tested hue recall (all samples and test responses were designed to vary in hue and minimally in luminance and saturation).

This is one possible criticism of the studies. One of the major improvements in Delta E 2000 is a hue dependent rotation. We are more or less sensitive to certain hue regions. While this is one source of criticism, they are generally relying on large color differences to differentiate successful trials vs. unsuccessful trials and Delta E does not have any meaning for large color differences.

A more accurate reinterpretation might be to say that they aren't calculating color differences at all but rather the variation of the response distribution and the underlying probability density functions. The variation analysis is not dependent on individual color difference calculations.