How to run the Cicero discrimination experiment
This experiment uses:
• headphones: use the closed-back headphones.
1. Explain the experiment
Script: "This experiment has three parts. First, you will be speaking into this microphone to make recordings of your voice. Then I will pick some of those recording s for the third stage of the experiment. This will take about ten minutes. Finally, there will be a listening section, where you will be asked to press a button on the keyboard based on what you hear. I'll explain the instructions for that part when we get there.
For now, we'll start with the speaking section. You will see one word at a time appear on the screen. It's a small number of words; they will repeat themselves randomly. First you'll get to practice to get used to the feedback that you will receive about the duration of your vowels. You want to try to hold out the vowel to the appropriate length. After you speak, you'll see a horizontal rectangular bar indicating the approximate length of your vowel. You're aiming for a half-full bar. If it was the right length, the feedback bar will turn green. If it is almost right, it will be yellow, and if it's too short or too long it will be red. You'll get a chance to try again on the next trial. The duration feedback will continue to display throughout the experiment. There will be breaks at regular intervals."
2. Run the recording experiment (Part 1)
First, run check_audioLevels as per usual.
IF SUBJECT HAS COCHLEAR IMPLANTS (CIs):
For most people, the mic that is one thicker bar with a kind of node near the temple is a better fit and gets less tangled in a CI. I demo how to put it on myself (it's kind of confusing!) and then present the mic to them in the orientation that the mic should be in for them to put it on. Other people may prefer the other mic. Listen to their preference, and then adjust the mic as usual.
Some people who wear cochlear implants use over-ear headphones, especially younger ones or people who may use headphones at their job. I typically ask if they have ever used over-ear headphones before, and if they have, then I give them the headphones and let them adjust on their own head. Don't forget that their sound input is going to be higher (above the ear) than someone without a device.
Audio equipment is likely to change or to have changed since writing this. Check for playback volume level in the discrimination task, which uses PsychToolbox and may need some fine-tuning. If you are new to the lab (welcome!) or haven't done much with PsychToolbox audio, you can get a longer-timer to help you with this. If this is totally opaque to everyone, Sarah Bakst will be happy to help troubleshoot from afar.
In the command line, type: speak_only
To do: make wrapper for this and call it something else.
A. Complete the training phase
a. There is duration training, and participants will receive duration feedback during trials.
b. Let the training run. If people can't get their utterances long enough, remind them that it's only the vowel that he machine is tracking, or that speaking louder is not the objective.
c. The training will ask if you want to re-run. If successful on less than 13/15, rerun.
B. Continue with full experiment
When training has been satisfied, move on to the full experiment. At the end of the experiment, hit spacebar. There will be a pause as Matlab collects the data, and then it will finish.
3. Run the discrimination experiment (Part 2)
This experiment is divided in three parts (like Gaul
1. Choosing specimens from words the subject has already produced
2. Making stimuli from those specimens
3. Discrimination task
1. Choosing specimens.
You will pick words for this experiment! Ask the subject to wait while you set up this part of the experiment. You can tell them that it will take a few minutes, and they're welcome to take a little break.
Put on headphones.
Load in the experiment (double click expt.mat) from the subject you have JUST run. It will be in the experiments folder under cicero/spXXX/production/.
at the command line, run:
First, you will hear five instances of the word Ed, then Matlab will ask you which of the five sounded good to you. You have the option to hear them all again, or if they were all bad, to hear five new ones.
Criteria for good stimuli:
You're listening for stimuli where the subject
• was not yawning, sneezing, sniffling, mouth-smacking, or producing some other interference
• did not have creaky voice (media calls this "vocal fry," but the real name is "creak"), if possible
• did not have a super-short utterance
• sounded normal for that subject (i.e. the stimulus didn't sound super different from the other four)
It's entirely possible that more than one of those five (or none of those five!) fit the criteria. If more than one of them sounds good, just pick one.
After you choose the Ed stimulus, there will be a pause, and you'll start hearing oh utterances. Same criteria apply.
2. Making stimuli (+ audiometric thresholds, if not CI).
At the command line, run:
If subject has CIs:
If subject has normal hearing:
IF SUBJECT DOES NOT HAVE CIs:
Now is a great time to do the audiometric thresholds!
Once the code has finished running, ask the participant to sit and put on the headphones. You'll want to use the closed-back headphones (the ones without the holes).
Test to make sure the volume is comfortable:
^ I made this script in March 2020 so it has not been extensively tested. Be sure to test before running subjects again.
IF SUBJECT HAS CIs:
See above about headphones!
3. Running discrimination experiment (twice!).
At the command line, run:
To do: make wrapper for this and edit here with the new name.
First there will be a training phase, where the subjects will figure out how to respond during the experiment. Subjects will be given the chance to ask you any questions or to hit spacebar to move on. After they hit space, you also will also need to hit space. This is to give you control over when the experiment moves on. If you're ready, hit spacebar.
If you need to quit, hit ctrl + c until either the experiment looks like it is stopping or you get red error messages. I still haven't figured out what the bug is, with re-running the practice. Maybe we should just have a separate practice module.
Type sca at the command line.
If you need to rerun the practice, quit (ctrl + c) and try again.
After they make it through, run the same command again. They'll be doing this experiment once for Ed and once for oh.
Typically, subjects will then move on to cif1
(pronounced sci-fi, haha!)