PsychAdapt: A Method for Maintaining Performance During Perception Tasks
Often times in perception research we may wish to have participants respond to stimuli with a certain level of accuracy. Typically, most experimenters will use one of various methods (Weibull function, logistic function, etc.) to estimate a participant’s threshold at a given detection level, and stop the estimation after a certain criteria has been met (i.e number of reversals).
For example, one may choose to estimate a participant’s threshold of detection or discrimination performance at 75%, and then continuously present stimuli at that level in order to fit the needs of the experiment. However, the predicted stimulus threshold at 75% may not remain constant. This can be due to a variety of reasons such as participants becoming more familiar with a task, or fatigue effects that may arise over the course of a lengthy perception experiment.
Therefore, simply estimating a threshold for a given level of performance and testing at that level is not good enough. The task must be able to adapt to changes in participant performance for the duration of the experiment. Perceptual adaptation methods do exist, and are routinely used in sensation and perception research. However, many of these methods were developed to estimate a participant’s threshold for a given stimulus, but not necessarily maintain a continuous level of difficulty. These methods, such as PEST, staircase procedures, and QUEST are all reasonable, and proven methods at estimating stimulus thresholds for various tasks. They just weren’t designed to maintain task performance for the entirety of an experiment (as measured by % accurate answers).
PsychAdapt was designed specifically to estimate stimulus thresholds at any level of accuracy using a logistic psychometric function, and crucially, to maintain that level of accuracy by continuously sampling at points around the predicted threshold level. Another key aspect of PsychAdapt is that it updates the logistic model after every trial therefore continuously adapting to changes in a participant’s performance while still maintaining a given level of accuracy, and therefore, average task difficulty.
If you’re interested in using PsychAdapt, or want to try it out, just download it from my Github page (linked below) and see what you can do with it. Keep in mind that PsychAdapt is currently under active development and the code may change rapidly or in unexpected ways. I’ll try to keep things from breaking as much as possible though.
