Severe bug in Inquisit Web: wrong images shown, correct images logged in data file


Severe bug in Inquisit Web: wrong images shown, correct images logged...
Author
Message
Blackadder
Blackadder
Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)
Group: Forum Members
Posts: 280, Visits: 147
We just discovered a severe error in Inquisit Web. The bug occurred with this script: http://research.millisecond.com/persike/simcomparegs.web

The script is supposed to show faces compsed from two face halves. In total, there are 200 lower and 200 upper face halves. The names of the image files uploaded to the Inquisit server accordingly have the prefix "UPPER" or "LOWER".

Multiple subjects have mentioned the same problem while running the experiment from their web browser. On their machines, stimuli are not composed from one upper and one lower face half. Instead, Inquisit shows two bottom halves.

We can definitely exclude that this is due to a script error. The data files logged by Inquisit Web contain the correct filenames (i.e., 1x "UPPER", 1x "LOWER") while only two "LOWER" variants are actually displayed. Below I have posted an image from one of our subjects documenting the bug, and attached you find a data file containing data from this exact experimental run.



This needs to be fixed.

Kind regards, Malte


Attachments
Dave
Dave
Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)
Group: Administrators
Posts: 12K, Visits: 98K
Blackadder - Wednesday, February 8, 2017
We just discovered a severe error in Inquisit Web. The bug occurred with this script: http://research.millisecond.com/persike/simcomparegs.web

The script is supposed to show faces compsed from two face halves. In total, there are 200 lower and 200 upper face halves. The names of the image files uploaded to the Inquisit server accordingly have the prefix "UPPER" or "LOWER".

Multiple subjects have mentioned the same problem while running the experiment from their web browser. On their machines, stimuli are not composed from one upper and one lower face half. Instead, Inquisit shows two bottom halves.

We can definitely exclude that this is due to a script error. The data files logged by Inquisit Web contain the correct filenames (i.e., 1x "UPPER", 1x "LOWER") while only two "LOWER" variants are actually displayed. Below I have posted an image from one of our subjects documenting the bug, and attached you find a data file containing data from this exact experimental run.



This needs to be fixed.

Kind regards, Malte


As weird as it may sound, but this may not actually be a bug in Inquisit, but a bug in the graphics card driver on the affected machines. We have seen a few similar instances over the years. If you can get one of the affected participants to update the graphics card driver to the latest one available from the respective vendor and then try again, that should either confirm or disconfirm the theory.

Blackadder
Blackadder
Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)Supreme Being (26K reputation)
Group: Forum Members
Posts: 280, Visits: 147
Dave - Wednesday, February 8, 2017
Blackadder - Wednesday, February 8, 2017
We just discovered a severe error in Inquisit Web. The bug occurred with this script: http://research.millisecond.com/persike/simcomparegs.web

The script is supposed to show faces compsed from two face halves. In total, there are 200 lower and 200 upper face halves. The names of the image files uploaded to the Inquisit server accordingly have the prefix "UPPER" or "LOWER".

Multiple subjects have mentioned the same problem while running the experiment from their web browser. On their machines, stimuli are not composed from one upper and one lower face half. Instead, Inquisit shows two bottom halves.

We can definitely exclude that this is due to a script error. The data files logged by Inquisit Web contain the correct filenames (i.e., 1x "UPPER", 1x "LOWER") while only two "LOWER" variants are actually displayed. Below I have posted an image from one of our subjects documenting the bug, and attached you find a data file containing data from this exact experimental run.



This needs to be fixed.

Kind regards, Malte


As weird as it may sound, but this may not actually be a bug in Inquisit, but a bug in the graphics card driver on the affected machines. We have seen a few similar instances over the years. If you can get one of the affected participants to update the graphics card driver to the latest one available from the respective vendor and then try again, that should either confirm or disconfirm the theory.

Interesting. We'll try that.

Thanks a lot!
Dave
Dave
Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)Supreme Being (1M reputation)
Group: Administrators
Posts: 12K, Visits: 98K
Blackadder - Wednesday, February 8, 2017
Dave - Wednesday, February 8, 2017
Blackadder - Wednesday, February 8, 2017
We just discovered a severe error in Inquisit Web. The bug occurred with this script: http://research.millisecond.com/persike/simcomparegs.web

The script is supposed to show faces compsed from two face halves. In total, there are 200 lower and 200 upper face halves. The names of the image files uploaded to the Inquisit server accordingly have the prefix "UPPER" or "LOWER".

Multiple subjects have mentioned the same problem while running the experiment from their web browser. On their machines, stimuli are not composed from one upper and one lower face half. Instead, Inquisit shows two bottom halves.

We can definitely exclude that this is due to a script error. The data files logged by Inquisit Web contain the correct filenames (i.e., 1x "UPPER", 1x "LOWER") while only two "LOWER" variants are actually displayed. Below I have posted an image from one of our subjects documenting the bug, and attached you find a data file containing data from this exact experimental run.



This needs to be fixed.

Kind regards, Malte


As weird as it may sound, but this may not actually be a bug in Inquisit, but a bug in the graphics card driver on the affected machines. We have seen a few similar instances over the years. If you can get one of the affected participants to update the graphics card driver to the latest one available from the respective vendor and then try again, that should either confirm or disconfirm the theory.

Interesting. We'll try that.

Thanks a lot!

Please let me know what you find.

To illustrate why I think this may be what's happening, here's one relatively recent example of a similar issue caused by a faulty graphics card driver:

-------- Forwarded Message --------
Subject:     Re: Problem with script
Date:     Wed, 20 Apr 2016 13:29:07 +1000
From:     [EMAIL REDACTED]
To:     support@millisecond.com

Hi David,

Thank you for the advice. It worked. Excellent work!

Regards,
[NAME REDACTED]

--
[NAME REDACTED], PhD
Research Fellow
[INSTITUTION REDACTED]
--

On Wed, Apr 20, 2016 at 12:31 PM, Support <support@millisecond.com <mailto:support@millisecond.com>> wrote:

    Hi [NAME REDACTED],

    The problem is almost certainly caused by a faulty graphics card
    driver on the affected system, not by an issue in the script or in
    the Inquisit application. We have seen a (small) number of similar
    cases over the years. In all of these cases updating the graphics
    card driver to the latest version available from the card's
    manufacturer resolved the problem.

    Hope this helps,
    -David

    On 4/20/2016 3:28 AM, [NAME REDACTED] wrote:

       Hi Support

       I experience the following problem:
       When I run the attached script on he computer I am writing from
       now it does not work as expected. The script is a slightly    
       modified version of a script from your website. It works
       correctly on some other computers.
       The problem is that the script should show 5 fishes where the
       middle one in 50% of cases should look at the opposite direction
       compared to the others. It never happens on this computer. All
       fish always look at the same direction on screen although when I
       press the correct key according to what I see the program
       reports an incorrect input. The code and the log are attached.
       Please advise what should I do to fix it.

       Regards,
       [NAME REDACTED]

       --
       [NAME REDACTED], PhD
       Research Fellow

       [INSTITUTION REDACTED]
       --

GO

Merge Selected

Merge into selected topic...



Merge into merge target...



Merge into a specific topic ID...




Reading This Topic

Explore
Messages
Mentions
Search