[open] Acquiring Unwanted Frames.

booradley

EDIT 1: Just remembered I started a related thread previously. Here is the link which gives some detail on the issues then. We were getting 6% error in trials then and now 20%.

http://www.cogsci.nl/forum/index.php?p=/discussion/346/open-visual-presentation-times-vary-massively-and-regularly-#Item_13

EDIT 2: I made a few tweaks and got errors down to 6% error rate again. However, ideally we'd still like to reduce that too.

Hi folks,

I'm currently trying to run a follow up study.

The following seems to be a consistent problem I've had across studies.

When running through trials and measuring how many frames are being used per Sketchpad (when Sketchpad is set to '1' to ensure 10ms display on a 100Hz monitor) I seem to be acquiring extra frames randomly.

We've used both an oscilloscope to record this and also calculated in the CSV files which match the oscilloscope, generally.

The previous experiment added unwanted frames at around 6% of trials, however, the current on is around 20%.

The only difference in paradigms is where I place inline scripts. Literally the same number and type of items are used.

Below are some screen shots that might help with getting a feel for the experiment.

LOOP used:

RUN ifs in SEQUENCE:

SEQUENCE:

WARNING MESSAGES:

I've looked at the CSV output and there does not seem to be a pattern to where the extra frames are acquired.

As with the previous experiments we have achieved accuracy with the presentation of external beeps but this last hurdle is a real thorn in our side as the next experiment is the one where we need 100% accuracy in order to support our findings and claims.

Any help would be appreciated.

Thanks,

Boo.

sebastiaan

Hi Boo,

Looking at your experimental_sequence, I think there may be too many stimuli to be preloaded. When preloading fails, OpenSesame (or actually PsychoPy, assuming you are using the psycho back-end?) will sometimes generate the stimuli at runtime (i.e. when they are presented), and this will lead to unpredictable delays.

There are a few things that you could do.

• Firstly, you could use feedback items for all the non-time-critical displays. feedback items are not preloaded, so using these should reduce the load on your system.
• Secondly, you can use 'Show if' statements to determine which elements are displayed on a specific sketchpad. This is more efficient than creating multiple sketchpads and selecting one at runtime using a 'Run if' statement. But in many cases it can accomplish the same thing.

Does this help at all? I think the problem is basically a system overload, and can be resolved by implementing the experiment a bit more efficiently.

Cheers!
Sebastiaan

booradley

Hi Sebastiaan,

That's pretty neat. I'll be sure to give these a go.

Thanks yet again for all your help.

Boo.