repeat practice trials after incorrect response

hi,
during the practice phase of my experiment, if the subject responds incorrectly, i want to show an "error" message, and then repeat the same trial again and again until he/she responds correctly.

the only way i can think to implement this, is to jump from the practice trial (#2) to a trial showing the "error" feedback (#100) and from there back to the practice trial (#2) again.

this doesn't work. the jump from the error trial back to the practice trial doesn't happen. maybe because it's a backwards jump (from trial 100 to trial 2) or maybe because it's a "double" jump, i don't know. i tried using both the jump and the skip functions, but neither works.
i can't use the incorrect.txt solution because it will interfere with the experiment phase (after the practice phase)

any ideas?

thanks,
michal

Currently, you can't jump backwards. Here is another way you can do it though..

Have duplicate trials. e.g., trial 1, 2, 3, 4, where 1 & 2 are identical and 3 & 4 are identical. If the respondent answers #1 correctly, they skip to #3 (the "real" next trial), otherwise, they go to #2 (a repetition of #1). Effectively, this repeats the trials *and* records the data for the failed first attempt. You can apply this logic indefinitely depending on how many attempt you want to allow. Finally, on the repeated trials, you could add an initial stimulus indicating that they were incorrect.

Note you could also do this *within* trials.. repeating stim, loc, time sets and skipping them out of the trial only when they respond with the correct answer, e.g., rt:57>2!,43

[COLOR=darkred][correction 9/12/07: the syntax [I]rt:57>2!,43[/I] should be [I]rt:57>-2,43. [/I]see editor's note below][/COLOR]

Does that make sense?

I ran into the same problem with trying to backwards jump trials. I wish this were possible - esp. when you want subjects to be able to repeat the trial as many times as is necessary to submit the correct answer. Theoretically, you could just keep inserting identical trials, one after the other, but there is also a limit to how many trials direct RT can hold (99, I think?). [COLOR=darkred][nope--editor's note see below][/COLOR]

But simply inserting identical trials does work.

One question for Jarvis24. What does the "!" after the "2" mean? I am unclear what "rt:57>2!,43", as opposed to just "rt:57>2,43" would do. [COLOR=#8b0000][yikes! see editor's note see below][/COLOR]

OK, that's really weird. I don't know why I would have written it that way above (i.e. using the ! symbol). Unless I'm too tired to be thinking about this, I'm sure it should be [I]rt:57>-2,43[/I] if you want to jump out of the trial immediately on the pressing of the spacebar (as opposed to waiting until the whole trial is finished). I don't know where the ! came from. It may be that at one point I considered that as a possible syntax for the function. In any case, I don't think it would have any such power if you were to try it. And hey, if you do try it, let me know!

Also, just a note of clarification. There [I]is[/I] a limit in DirectRT of 99, but it's a limit on the number of stimuli [I]within a single trial[/I], not a limit on the number of trials. I'm sure that's what you meant, but I wanted to be super clear for the sake of all those following along at home :)

[quote=jarvis24;1765]Currently, you can't jump backwards.[/quote]
In other words, you cannot do a loop in DirectRT.

(I added this comment to this thread just so that people like me who want to find out whether DirectRT can do loops can now find it by searching the forum with the terms "loop" or "loops".)

-- David McFarlane, Professional Faultfinder

Thanks David. Btw, it might be fun to know that you can also use [I]Google[/I] to search these forums also. For example, I was able to find this thread now by Googling: [I]directrt loop[/I]

[I]e.g., [URL]http://www.google.com/search?q=directrt+loop[/URL][/I]

You can always add [I]empirisoft[/I] or [I]empirisoft support[/I] to be sure you get the support forums (e.g., if you end up getting hits from MIT when searching on [I]MediaLab[/I]).