During the experiment, participants sat in front of a computer screen in a sound-attenuated room while stimuli were played binaurally through a headset. Participants read instructions presented on screen that described the ABX paradigm. They were told that in every trial, the first two nonwords they heard would be different and that the third nonword they heard would always correspond to either the first or second nonword. They were asked to give their response by pressing one of two buttons on a button response box.
In a given trial, participants heard three different stimuli, each produced by one of the three synthetic voices (voice order was randomized across trials). An ISI of 300 ms was used between both the A and B and the B and X tokens. After the offset of the X token, participants had 2,000 ms to give a response. An additional 1,250 ms was given to participants between each trial.
Twelve counterbalancing groups were created to distribute the 1,728 unique contrast~position~surrounding-consonant~vowel~presentation-order combinations into 144-trial long lists, such that each subject heard six trials for all twenty-four one-feature obstruent contrasts, with two trials for each vowel /a, i, u/.
Half of the groups heard each unique trial in A-B order, and the other half in B-A order, though all subjects heard two versions of each trial (one where the correct response was A and one where the correct response was B).