Consider a psychophysics procedure using a standard staircase method to adaptively find the threshold, using a 2AFC paradigm and a 1-up, 3-down method to determine the 75% threshold. Now suppose the use of steps that are unequal in size. I am not referring to decreasing the step size after a certain amount of reversals as described by Levitt (1971), but to the use of unequal step sizes. So for example suppose that the available steps are 0.5 - 0.5 - 0.5 - 1 - 1.5 - 0.5 - 2.5 - 2. So this would translate to a discrete set of available stimulus levels of: 0 - 0.5 - 1 - 1.5 - 2.5 - 4 - 4.5 - 7 - 9. Since step size can be used to alter the correct rate at which the staircase procedure converges (e.g., 50%, 75% etc.) Kaernbach (1991), I was wondering what my proposed stimulus-level distribution would mean for the threshold definition of a 1-up 3-down procedure intended to converge onto a 75% threshold?


It is going to mean bad things. If you can only create stimuli with a predefined non uniformly sampled set of "levels" then the the corresponding percent correct that the staircase will track will depend on the threshold of an individual observer. For example, if you can create stimuli with levels of 0, 1, 2, 3, ..., 100, 110, 130, 170, 250, 410, where for small levels the up and down step sizes are equal, but for large levels the up step size is twice the down step size. This means the percent correct point is going to be different depending on the threshold of the subject. You wouldn't have to worry about it if you use something like the method of constant stimuli.

| improve this answer | |
  • $\begingroup$ But basically the staircase method bases the threshold on reversals. A reversal is simply the mean of two stimulus levels where the answer changed from + to - or vice versa. Isn't this method then robust to step level? In addition, sound levels are generally measured in log units whereas distances are often used in absolute, non-transformed measures. Wouldn't the use of log units (dBs) therefore mean bad things as well? $\endgroup$ – AliceD Nov 5 '14 at 11:56
  • 1
    $\begingroup$ @ChrisStronks Kaernbach (1991) says no. As for linear versus log scale. Yes, a 3-down 1-up track on sound pressure measured in linear units would not converge to the same point as a 3-down, 1-up track on sound level measured in decibels. most auditory researchers care about decibels, so they use decibels. $\endgroup$ – StrongBad Nov 5 '14 at 12:03
  • $\begingroup$ just to clarify: to what says Kaernbach (1991) 'no' to exactly? Would you have additional references for me? $\endgroup$ – AliceD Nov 5 '14 at 22:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.