Purchase this article with an account.
Matjaz Jogan, Alan Stocker; A new 2AFC method for the comparison of stimuli that differ along multiple stimulus dimensions. Journal of Vision 2012;12(9):619. doi: 10.1167/12.9.619.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
2AFC methods are frequently used to assess the influence of a secondary stimulus parameter (e.g. stimulus noise) on the percept of a primary stimulus dimension (e.g. motion direction). In this case, test and reference not only differ along the primary stimulus dimension, but also in their value of the secondary stimulus parameter. As a result, even a thorough instruction does not guarantee that a subject's choice is exclusively based on the perceived differences along the primary stimulus dimension. This can lead to an incorrect characterization of perceptual biases and thresholds.
We present a new 2AFC method that prevents this problem by using two reference stimuli instead of one. On each trial, the test and the two reference stimuli are simultaneously presented, and the subject has to choose the reference that is perceptually closest to the test. Because the choice is between the two references (as opposed to reference and test), the decision can only be based on the perceived difference along the primary stimulus dimension, thus avoiding any potential confounds in the decision process.
We can formulate an observer model for our new method based on standard signal detection theory. It allows us to extract the point-of-subjective-equality as well as the noise widths for both references and test independently. An adaptive staircase procedure based on a continuous entropy measure allows the efficient allocation of trials within the experimental parameter range. Numerical tests confirmed both, convergence and efficiency of the new method. Finally, we applied the new method to a random-dots motion discrimination experiment, characterizing perceived motion direction (primary dimension) for different dot coherence levels (secondary parameter). Subjects showed no problems to perform the task. Model fits well capture subjects behaviors, and the extracted noise widths correctly reflect the coherence levels, thus demonstrating the practicability and validity of our new method.
Meeting abstract presented at VSS 2012
This PDF is available to Subscribers Only