Just Noticeable Difference

Scientific calculation of the average detectable difference between two measurable qualities, such as weight, brightness of light, loudness of sound.

When we try to compare two different objects to see if they are the same or different on some dimension (e.g., weight), the difference between the two that is barely big enough to be noticed is called the just noticeable difference (JND). Just noticeable differences have been studied for many dimensions (e.g., brightness of lights, loudness of sounds, weight, line length, and others).

The human sensory system does not respond identically to the same stimuli on different occasions. As a result, if an individual attempted to identify whether two objects were of the same or different weight he or she might detect a difference on one occasion but will fail to notice it on another occasion. Psychologists calculate the just noticeable difference as an average detectable difference across a large number of trials. The JND does not stay the same when the magnitude of the stimuli change. In assessing heaviness, for example, the difference between two stimuli of 10 and 11 grams could be detected, but we would not be able to detect the difference between 100 and 101 grams. As the magnitude of the stimuli grow, we need a larger actual difference for detection. The percentage of change remains constant in general. To detect the difference in heaviness, one stimulus would have to be approximately 2 percent heavier than the other; otherwise, we will not be able to spot the difference.

Psychologists refer to the percentages that describe the JND as Weber fractions, named after Ernst Weber (1795-1878), a German physiologist whose pioneering research on sensation had a great impact on psychological studies. For example, humans require a 4.8% change in loudness to detect a change; a 7.9% change in brightness is necessary. These values will differ from one person to the next, and from one occasion to the next. However, they do represent generally accurate values.