(*instead of using Timeout)

You should use either /timeout or /trialduration, set to a random value below 3 seconds.

<values>
/ randomduration = 0
</values>

<sound mysound>
/ items = ("threeseconds.wav")
/ playthrough = false
</sound>

<trial sometrial>
/ ontrialbegin = [values.randomduration = round(rand(10,2999))]
/ stimulusframes = [1=mysound]
/ validresponse = (0)
/ timeout = values.randomduration
</trial>

There is no other way to "stop" a sound once it's started playing.

Thank you for your advice, Dave.
Best,
nakayama 

Hi, Dave.
It is possible to let this random value drawn following a normal distribution with a certain fixed mean and standard deviation, say, 2000 and 40? So the timeout stop approximately 68% times within 2000+_40.
ontrialbegin = [values.randomduration = round(rand(10,2999))]
Best,
Nakayama

I am trying to randomize the onset (stimulustimes) of the stimulus using this syntax, but it seems not working.

<trial time>
/ ontrialbegin = [clear(trial.time.stimulustimes]
/ ontrialbegin = [setitem(trial.time.stimulustimes,shape.targetdot,list.targetduration.nextvalue)]
/ stimulustimes = [1=fixation;2000=targetdot]
/ timeout = 2900
</trial>
Is there a way to fix this?
Thank you.

Regarding your first question (sampling values from a normal distribution), see https://www.millisecond.com/forums/Topic8431.aspx

Regarding randomizing the onset of the stimulus, your syntax is invalid. You need to use the trial elements setstimulustime() function. See e.g. https://www.millisecond.com/forums/Topic11671.aspx as well as the language reference for setstimulustime(), resetstimulusframes(), etc.

Thank you Dave. Very appreciated for your advice.
The one for normal distribution generation appears complicated to comprehend so I decide to use discrete stimuli to approximate Gaussian.
Looking forward Inquisit to have some functions like norm().
Best,
nakayama

Hi, Dave.
Sorry for asking again.
Following your advice, I made this:
However, onset of items appears synchronous, is there a way to varying them using the values from list?

<list targetduration>
/ items = (2100,2200,2300,2400,2500)
/ selectionmode=random
</list>

<trial timepresent1>
/ ontrialbegin = [trial.timepresent1.setstimulustime(shape.dot2,list.targetduration.nextvalue);
trial.timepresent1.insertstimulustime(shape.dot3,list.targetduration.nextvalue);
trial.timepresent1.insertstimulustime(shape.dot4,list.targetduration.nextvalue)]
/ stimulustimes = [0=fixation;2000=erase]
/ timeout = 2900
/ ontrialend = [trial.timepresent1.resetstimulusframes()]
</trial>

If you need to sample more than one value from a <list> in a given <trial>, you need to set its /selectionrate to always. The default is trial. Details can be found in the documentation for the /selectionrate attribute.

<list targetduration>
/ items = (2100,2200,2300,2400,2500)
/ selectionmode=random
/ selectionrate = always
</list>

You'll also have to think about resetting the list before it runs out of items (you are sampling without replacement). You'll want to do

<trial timepresent1>
...
/ ontrialend = [trial.timepresent1.resetstimulusframes(); list.targetduration.reset(); ]
</trial>