Voice-recorded picture naming

Collect spoken picture-naming responses, saving an audio recording and the voice onset for every trial.

When to use it

Use this when the response of interest is speech: picture naming, word reading, verbal fluency probes, or any task where naming latency and the utterance itself both matter. A keyboard proxy (press space when you start speaking) is unreliable and loses the utterance entirely, while the microphone gives you the onset latency and the audio to score offline. If you only need to know that the participant spoke, not what they said, a plain voice key with nothing saved is lighter on storage.

Trial file

type	stimFormat	stim1	responseType	responseOptions	responseWindow
test	.jpg	dog	mic	save:2000;prevoice:200	6000
test	.jpg	broom	mic	save:2000;prevoice:200	6000
test	.jpg	apple	mic	save:2000;prevoice:200	6000
test	.jpg	bicycle	mic	save:2000;prevoice:200	6000
test	.jpg	candle	mic	save:2000;prevoice:200	6000
test	.jpg	ladder	mic	save:2000;prevoice:200	6000

How it works

responseType set to mic turns the trial into a spoken-response trial. Using mic anywhere in the column also switches on microphone setup for the whole session: participants grant mic access and read a short calibration passage aloud, which sets an individual voice-detection threshold before the first trial.

On its own, mic is only a voice key: speech onset advances the trial and nothing is kept. The responseOptions cell is what saves audio. save:2000 records from the detected voice onset until onset plus 2000 ms, then the trial advances on its own. prevoice:200 keeps the 200 ms of audio just before the detected onset, so soft consonants at the start of a word are not clipped; it must sit in its own semicolon-separated segment, as here. Unlike most responseOptions flags, the mic flags are not case-sensitive. Because this trial saves its recording, the picture stays on screen while the participant speaks; only unsaved voice-key trials hide the stimuli at the response.

The numeric responseWindow of 6000 is the guard for silent trials: with onset-anchored saving, a trial where no speech is detected would otherwise wait forever. After 6000 ms the response is recorded as timeout and the trial moves on. On mic trials a timeout that lands during a live recording is held until the recording is finished, so audio is never cut off mid-save.

Each saving trial adds audio columns to the results, including audioFile (the saved file, named after the trial row, for example row3.wav) and voiceOnset (when the voice was first detected, in ms). Together with recOnset, the difference voiceOnset - recOnset gives the exact position of the onset inside the audio file. There is no key cell here, so the correct column stays empty; naming accuracy is scored offline from the recordings.

Variations

• Record the whole trial instead of an onset-anchored clip with responseOptions set to save:all, but always pair it with a button, a keyboard key, or a responseWindow deadline, because voice alone never ends a save:all trial.
• Keep the onset-anchored duration but capture from trial start by adding start:trial in responseOptions, for example save:2000;start:trial.
• Strip it down to a pure voice key by leaving responseOptions empty: onset advances the trial, nothing is saved, and the stimulus is hidden at the response unless you add noHide.
• Hide the pulsating recording indicator with hideRec in responseOptions when the on-screen REC cue would distract participants.