[solved] Freeze after around 450 items

We have a long experiment, around 650 response scale items, and MediaLab freezes after around 450 items or so. It has never gone through all of them. The background screen appears, but no item appears. It just stays like that. MediaLab handles each individual questionnaire fine (the longest is 240 items). Basically, it resembles the old "crash at 150" bug that was fixed in November and has been extensively discussed here.

The specs: MediaLab has failed on both Windows XP SP2 and Windows 7 Enterprise. Both computers have images less than 2 weeks old, and both have a version of MediaLab that was downloaded last week.

Any tips?

It's odd that it would happen on XP. Could you try this please--if you create a second version of the study where the second half of the items comes first, does that alter when the error occurs? i.e., does the crash happen earlier?

The order doesn't seem to matter: reversing the order of questionnaires within the experiment doesn't fix the freeze, but reducing the number of items does. (I can send the "Test Experiment" I developed back-channel, but the scales can't be posted in a public forum.)

It is indeed odd to find this on a stripped-down XP SP2 machine with basically nothing on it as well as on a Windows 7 Enterprise machine that's stuffed with software.

Are any other ML users having success with long experiments?

Yes, please do zip and send me the experiment folder if you can (support@empirisoft is fine if I know it's coming). I'll run it here and see what I can do. There were memory leak issues with large experiments on some systems due, we think, to a file called evideo.ocx. However, we are just wrapping up v2010 this week and have removed this file completely now. It was an essential component of the program but we've re-written the code to be much more efficient and *should* solve these experiment-size oriented issues. But before I make any promises, please let me take a look at what you've got and I'll see if I can run it successfully here--both on XP and 7.

Here's an update on the freezes in the most recent ML 2008.

The short version: the ML 2010 beta version I downloaded worked exactly right.

The long version:

I ran the experiments in ML 2010 beta on both a Dell Latitude E5500 (dual 2.40 GHz, 4 GB RAM, 32 bit) running a fresh Win 7 Enterprise image (a few weeks old) and an ancient Dell Optiplex 260 (1.8 GHz, 2 GB RAM, 32 bit) running XP SP3 that has been updated but never re-imaged since 2002.

I ran both a test file (ca. 600 scale response items) as a real experiment file, which contains 8 different questionnaires with many fill-in-the-blank items, check lists, skipping and branching, instruction slides with durations, WGR for hundred of items, and so on. It doesn't have audio, video, or images, but it has around 650 to 700 variables.

There wasn't any hanging or freezing, and the SPSS files were correct for both machines and both experiments.

The 2010 beta build certainly fixed whatever vexing issue was causing the freezing in these experiments.

Thanks for the update Paul--very helpful. If any other users find this thread and are experiencing this or a related problem, let us know and we can get you the pre-release link for the 2010 version. Although it's technically in "pre-release" as of today (Jan 31st, 2011) it will be officially released on the website sometime this week. But again no need to wait if you want to try it now. Just email [EMAIL="sales@emp"]sales@emp[/EMAIL]...

[I]Note from moderator:[/I]

The "x" parameter item type was not working in this initial release of 2010. Requesting a "password to continue" in this manner would produce a blank page. This has since been fixed as of 2010.2.19, posted Feb 17, 2011.

See also Version History for v2010.2:
[URL]http://www.empirisoft.com/Support/showthread.php?t=5577[/URL]