The present dataset has been collected through an experiment on fast ITD discrimination thresholds determination. Each participant was sitting in front of the laptop computer running the GUI during the test. At each trial, the task consisted of listening to two subsequent and randomly balanced stimuli, and then choosing which sound was the rightmost, by selecting it with the mouse on the computer screen. At the beginning of each session, five pilot trials were presented having ITD levels equal to 240, 200, 160, 120, and 80 μs. Correct guesses in all such trials were necessary for the measurements to start in correspondence with the sixth trial. The session lasted approximately 10 minutes. The protocol was designed for determining the PF 79.4% threshold, describing the subjective lateralization performance as a function of T defined as half of the actual ITD to be discriminated. The target (i.e., rightmost) stimulus was lateralized twice as much as a nominal ITD, and the reference source (i.e., leftmost) stimulus was instead lateralized with an opposite ITD. Hence, an ITD threshold equal to 2T means that a participant discriminated the target ITD by T μs from the reference ITD of −T μs. The protocol implemented three different procedures: • adaptive 'three down, one up' two-interval 2AFC (2I-2AFC hereafter), • Gaussian Process Classification (GPC) with Bayesian Active learning by disagreement (BALD) (BALD hereafter), • GPC with random acquisition function (RANDOM hereafter). Accordingly, every session included three series of trials respectively implementing such procedures in a randomly balanced order. When GPC was used, thus enabling active learning and random selection of ITDs, the number of trials was empirically set to 15. In the dataset you will find: - Participant: IDs of the participant - Age: age of the participant - Gender: gender of the participant - ITD_2AFC: T values presented in the 2I-2AFC procedure -LABEL_2AFC: binary labels for the right (1) or wrong (0) answers in the 2I-2AFC procedure - PRED_2AFC: Weibull fitting predictions on 1-100 microseconds - ITD_BALD: T values presented in the BALD procedure - LABEL_BALD: binary labels for the right (1) or wrong (0) answers in the BALD procedure - MEANS_BALD: means of the GPC Bernoulli likelihood computed on 1-100 microseconds at each iteration step in the BALD procedure - VARS_BALD: variances of the GPC Bernoulli likelihood computed on 1-100 microseconds at each iteration step in the BALD procedure - MEANS_LATENT_BALD: means of the latent GPC prior computed on 1-100 microseconds at each iteration step in the BALD procedure - VARS_LATENT_BALD: variances of the latent GPC prior computed on 1-100 microseconds at each iteration step in the BALD procedure - ITD_RAND: T values presented in the RANDOM procedure - LABEL_RAND: binary labels for the right (1) or wrong (0) answers in the RANDOM procedure - MEANS_RAND: means of the GPC Bernoulli likelihood computed on 1-100 microseconds at each iteration step in the RANDOM procedure - VARS_RAND: variances of the GPC Bernoulli likelihood computed on 1-100 microseconds at each iteration step in the RANDOM procedure - MEANS_LATENT_RAND: means of the latent GPC prior computed on 1-100 microseconds at each iteration step in the RANDOM procedure - VARS_LATENT_RAND: variances of the latent GPC prior computed on 1-100 microseconds at each iteration step in the RANDOM procedure