In brightness matching, the brightness of one scene is varied by the experimenter until it matches the brightness of a second reference scene.
This paper demonstrated a tendency for test participants to set the illuminance of the varied scene to a slightly low illuminance. While the difference tends to be small, it can be sufficient to change a conclusion regarding the relative brightness of different scenes. This bias was demonstrated directly in the results of null condition trials, with an illuminance ratio (test/reference) of less than 1.0. The bias was also shown in the results of between-lamp trials. To overcome this, dimming should be applied to both the test and reference fields in successive trials.
This paper describes a bias revealed when a matching experiment was carried out at more than one reference illuminance (three in this case). Brightness settings were biased towards the middle illuminance, tending to be set low in trials with the high reference illuminance, and tending to be set high in trials with the low reference illuminance. This suggests a problem associated with the choice of reference illuminance (or other reference condition): it may be better instead to avoid choosing a reference and use all-possible-pairs.
This paper reviews the results of past studies using simultaneous (side by side) matching tests and suggests a need to:
Counterbalance the stimuli between the left and right hand sides to counter position bias
Apply dimming to both stimuli in turn
Set the variable stimulus to both high and low initial settings at the onset of trials.