Biases pose a major confound when inferring perception from behavior. Signal detection theory, a powerful theoretical framework for accounting for bias effects in binary choice detection tasks, cannot be applied, without fundamental modifications, to detection tasks with more than two alternatives. Here, we introduce a multidimensional signal detection model (the m-ADC model) for measuring perceptual sensitivity while accounting for choice bias in multialternative detection tasks. Our model successfully explains behaviors in diverse tasks and provides a powerful tool for decoupling the effects of sensitivity from those of bias in studies of perception, attention and decision-making that increasingly employ multialternative designs.
<br><br>
<b>Contents</b>:
<br> 1) Supplemental Data demonstrating key analytical results regarding the m-ADC model (Sridharan et al, J. Vis, 2014): Appendices E-F, Figures S1-S4 and Tables S1-S3.
<br> 2) Matlab scripts for maximum-likelihood and Markov-chain Monte-Carlo estimation of m-ADC model parameters (fit_mADC.m, MLE_4ADC.m).
<br><br>
<b>Update (September, 2014)</b>: Matlab scripts have been uploaded! The scripts are specifically for fitting four-alternative tasks (4-ADC tasks) [1,2]. The scripts can also be modified to fit a four-alternative forced-choice task (see fit_mADC.m, for instruction). If you would like to fit a task with a different number of alternatives (e.g., 2-ADC, 3-ADC, 5-ADC etc), please feel free to email the corresponding author at "dsridhar AT stanford DOT edu".
<br><br>
<b>References</b>:
<br> [1] Sridharan, D., Ramamurthy, D.L., and Knudsen, E.I. (2013). Spatial probability dynamically modulates visual target detection in chickens. PLoS One 8, e64136.
<br> [2] Steinmetz, N.A., and Moore, T. (2014). Eye movement preparation modulates neuronal responses in area V4 when dissociated from attentional demands. Neuron 83, 496-506.