[open] PyGame and RT recording

GuidoCor

Hi all.

I ran some experiments with PyGame legacy backend (I forgot to put in PsychoPy, my fault) and I have some problems retrieving the RT's information.

The experiment was like:
1- Show a cross
2- Show two pair of image on a sketchpad for ~300 ms
3- Show a mask until response
4- Keyboard response
5- Logger

The problem I found is that "flush pending keypresses" dind't work (but is selected...) . If a press the key in step 2, the mask appears and disappear and the recorded rt (response_time_choice) is sometimes 2, 3,4 ms depending on the machine that was ran. My question is: What about the other RT's? Are they reliable? My own hypotheses is that when you press the key in step two the RT is few miliseconds, but if you press in step 3 is the correct RT. I don't want to drop all the participantes who took the PyGame version, only the trials that have incorrect RT of 2,3,4.... : (

Am I right? Have someome had the same problem?

Thanks : )

sebastiaan

Hi Guido,

Yes, you're correct. Whether key presses are flushed or not is only relevant for keys that were pressed before the keyboard_item was called. So if you exclude RTs of say < 5 ms, then you should be fine in this respect.

You mention that flushing was enabled, but didn't work? Can you confirm this? And, if so, what version of OpenSesame and what operating system are you using?

Cheers!
Sebastiaan

GuidoCor

Hi, Sebastiaan. Thank you : )

We are running
Opensesame 2.9.7 on Ubuntu 14.04

Flush activated:

If you want I can upload some data set in which that 'error' ocurred.

sebastiaan

Hi Guido,

You're right, flushing key presses doesn't work on Ubuntu. Not as expected anyway, it works 'after a while'. This seems to be a bug in pygame, and it's specific to Linux and perhaps also Mac OS (I've heard similar complaints from Mac OS users).

Do you ever use pygame directly? If so, maybe you could dig around a bit to find a way to reliably flush the event buffer. What seems to help is to flush, wait 100 ms, and flush again. However, a 100 ms delay with every flush is hardly a desirable workaround. Quite annoying ...

Cheers,
Sebastiaan