N-back task

Comments

• eduard

  July 2021

  Hi Paula,

  I am not aware of a complete version, but I wouldn't be surprised if you stumble over on if you browse the forum or the web (e.g. OSF?). For example, here is a thread that had an (almost) working n-back task: https://forum.cogsci.nl/discussion/4389/n-back-task-problem There will probably be more.

  Hope this helps.

  Eduard

  
• Paula

  July 2021 edited July 2021

  Thank you for the reply!

  I am trying to create my own task relying on multiple examples and also your code in the linked thread. I'm happy to share it here once it works.

  I also want to log the 'false alarms', that is, when participants press the response button even if they should not. How can I do that?

  For now, I have a fixation of duration 0 after every stimulus. The challenge is counting a 'timeout' as a 'correct' answer... Edited: found a thread on this issue: https://forum.cogsci.nl/discussion/875/solved-set-timeout-as-correct-answer-on-a-go-nogo-task

  Do you have any thoughts on how to do that?
• eduard

  July 2021

  I also want to log the 'false alarms', that is, when participants press the response button even if they should not. How can I do that?

  You could code the correct_response per trial (None for nogos, and the response key for gos), and the actual response (var.response) per trial. If there is no response, it will be coded as None , so then it is as easy as comparing the correct_response and response variable and see whether they are not the same and the correct_response is None , et voila, you have your false alarms.

  
• Paula

  July 2021 edited July 2021

  I ran into a different challenge: how can I verify that (x-3)th letter in the first block is not the same as the xth of the second block?

  For example, for a 3 back task:

  if block 1 contains B, M, N, V, C, in this order, and

  if block 2 contains H, J, K, C, F in this order,

  then the C in block 1 and the C in block 2 are the same letter, but the programming (in the other thread) does not account for this...
• eduard

  August 2021

  Hi Paula,

  That depends on your implementation. But basically, you need to reset variables in every block. How did you deal with the first 3 trials in the first block? If what you did worked, do the same for the first three trials of every block. Does that make sense?

  If you need more specific help, please upload your experiment. It is easier to help you with your implementation, that demonstrating it with my own (which might be quite different than yours).

  Eduard

  
• Paula

  September 2021

  Thank you very much for the response! I managed to program the N-back task using different vectors, one that tracks the targets/non-targets and one that keeps track of the stimuli shown. Then, I compare the current stimulus with the stimulus that is situated N steps back in the stimulus vector. Well, it's this and a bunch of IF-statements for other possible outcomes :)

  Thank you for the help, @eduard !
• Paula

  September 2021

  Hi @eduard

  We decided to modify the Nback a bit and have stimuli equally spaced out at 2500 ms, regardless of whether they are a target or not. That is, if the correct answer is no response, then the timeout should be 2500 ms. If the correct answer is a response, the response will be given in whatever response time, and then there will be a blank sketchpad of 2500 - response_time. If I do this with a "coroutine", then I never actually get a timeout at 2500, but rather at 2498 or 2492 etc. Is there any workaround so that the timeout is exactly at 2500 ms? I upload the experiment below.

  Many thanks!

  nbacktask_forum - coroutine.osexp

  
  
• eduard

  September 2021

  Hi Paula,

  I took the freedom to change your experiment to work without coroutines. That way it becomes quite a bit easier to finetune the timing.

  nbacktask_forum - coroutine.osexp

  A small variability is normal. Have you seen the documentation page on timing? Normally, the timing should be fine like this, but it depends to some extent on your setup. For example, on my system with an external monitor attached, there was a jitter of 2496 - 2504. But if you run it on a single screen and in fullscreen, the timing should become better. So best, to check it on your experimental setup and evaluate whether the variability is acceptable or not.

  Hope this helps,

  Eduard

  
• Paula

  September 2021

  Hi @eduard ,

  Thank you very much for the change made! I greatly appreciate the help. With your workaround, I might also get it to work with the srbox, which is what we would prefer :)

  Best, Paula