ExperimentProcessDied Error
Hello!
We have been designing an experiment to run with OpenSesame but are unfortunately running into an error that is quite difficult to troubleshoot:
Error: ExperimentProcessDied
The experiment process seems to have died as the result of an unknown problem. This should not happen!
This error occurred in the unknown phase of item unknown.
Here’s what we’ve been able to figure out:
The crash typically occurs in the exact same place in the experiment. The experiment alternates between “exposure” loops and “test” loops. The exposure loops are identical to each other, and each time we present an exposure loop (20 trials with audio stimuli), we cycle through it 2.0x (total 40 trials). The crash almost always occurs on the second exposure loop, usually right at the 37th trial.
We know it’s not a specific stimulus file that’s causing the problem a) because the trials in each loop are presented in a random order and b) because an identical loop has successfully played out prior to the crash. We have tried making a simpler version (the current version dynamically selects a csv file with the right stimulus files based on experiment-wide parameters) but this error persists. We’ve also tried both loading the stimuli through a .csv file and through a table, but we get the same error.
We are suspecting this is a backend Python issue, which is also consistent with other posts we have seen on this forum. We are currently running pygame on OpenSesame version 4. We have also tried running the PsychoPy backend and get the same error message, but the crash happens earlier (in the first exposure loop) in that case.
We would welcome any suggestions on how to proceed with troubleshooting and are happy to share the OpenSesame file if that would be helpful.
Thanks!
Sahil
Comments
Hi @sahilluthra
Yes, this kind of error is often caused by a mismatch or instability in the underlying library versions and is kinda difficult to troubleshoot. Here are a few things you can try:
If none of these steps solve the problem, feel free to share your experiment file so I can test it locally and see if I can reproduce the crash.
But in some cases, this error can also stem from the audio backend or resource accumulation during long runs, so as a temporary workaround, you could also try splitting the experiment into two parts (conceptually two "blocks", in practice this would be two runs) to "reset" resources between blocks.
Let me know how it goes!
Claire
Check out SigmundAI.eu for our OpenSesame AI assistant! 🤖
Thanks, Claire! The experiment is indeed relatively long, so we opted to break it up into different runs following your suggestion. That seems to have taken care of the issue.
Best,
Sahil