-- KarlinBark - 29 July 2004
I worked on developing a JND test for skin stretch earlier this week. Although I talked about doing a comparison test of vibration vs. skin stretch earlier, I thought we hadn't explored the magnitude/skin stretch relationship enough. From our first testing, it appears that people can distinguish low, med, high forces with proper learning. It always seemed to take awhile for the subjects to realize what each force felt like, and then they had no problem detecting the various forces. The most difficult problem was being able to stretch the skin itself, because we were attempting to provide a purely tangential force on the arm. I think for a real device, there would be some small normal force applied to the arm in addition to the shear force, which would help prevent slipping. It may be difficult to apply what we designated a 'high' force, because the tape we used to stick to the arm would peel off easily.

Anyway, back to the JND test. The reasoning behind the completing a JND test was because during the magnitude testing, we realized that while the low force was easy to identify, the difference between the medium and high force was not as obvious. So whether or not this is something obvious, I don't know, but Li and I decided to find more concrete proof that the detection of skin stretch magnitude was not linear. For example, your skin is more sensitive to low force changes than high force changes, so you can feel the difference between 1 N and 0.9 N, but not between 5 N and 4.9 N. Well, that is the hypothesis.

So we devised a difference threshold test using the method of constant stimuli (Gescheider, 1997) and tested ourselves first. We used a reference stimulus of 1N and varied the comparison forces by 0.1 N so we compared 1N to 0.7, 0.8, 0.9, 1.0, 1.1, 1.2, and 1.3 N. We also tested varying forces with a reference stimulus of 5 N comparing it to 4.4, 4.6, 4.8, 5.0, 5.2, 5.4, and 5.6 N. The comparison values were chosen to make sure that the minimum comparison value and maximum comparison value would almost always be distinguished correctly from the reference stimulus.

The test compares each force to the reference stimulus 10 times, so there is a total of 70 comparisons. To get rid of time error (memory error), the reference stimulus is presented first in half of the comparisons, and second in the other half. The subject is given the two stimuli in succession and then asked which was the greater force.

We tested ourselves initially to get an idea of what the test would be like. The subject?s arm is upright, so that the spring scales that we use to pull on the subject's skin has as little friction and to negate any slack the spring scale has when lying on its side. The testing for each reference stimulus took about 30 minutes...the subject's arm becomes fatigued, as well as mentally fatigued. So in total, with the 2 reference stimuli, testing took approximately 1 hour.

The results from the testing show that there is likely a 0.2 N difference threshold for 1 N as a reference. The 5 N testing was less conclusive. The minimum and maximum comparisons were not low/high enough. Gescheider states that the min and max should almost always be distinguished correctly from the reference stimulus, but Li and I had difficulty distinguishing the min/max from the reference sometimes. So if we?re testing 5 N as a reference, the minimum and maximum should be less than 4.4 N and greater than 5.6 N. This also helps to validate our hypothesis that as the reference stimulus increases, the difference threshold will also increase.