[open] N-back task

Comments

• eduard

  February 2015 edited 5:19PM

  Hi Alex,

  However, the software, my own enthusiasm and limited Python coding skills are everything that I have, so far.

  This is usually enough, Python is quite easy to learn, so if you use your enthusiasm, you should be there in no time.

  Cannot find a way to make the trials fixed (e.g. instead of defining "n1" loop based on number of repetitions, it would be nice if the task was still self-paced, but the loop was terminated after, say, every 45 seconds).

  If you want it to be time-based, you can put the trial sequence into an inline_script in a way similar to this one:

  # the code gives you only pointers how to do it, it is not supposed to work by
  # copy & paste
  from openexp.canvas import canvas
  from openexp.keyboard import keyboard
  import random
  
  cv = canvas(exp)
  my_keyboard = keyboard(exp, keylist=['z', 'x'], timeout=3000)
  
  files_in_pool = ['star', 'circle', ...]
  
  
  # initialize timer
  t0, t1 = self.time(), self.time()
  while t1-t0 < 45000:
  
       # load image
       file = random.choice(files_in_pool)
       image_file = exp.get_file(file+'.png')
       stimuli.append(file)
  
       cv.image(image_file)
  
       # time response
       start_time = self.time() # as a reference for reaction time
       rp_key, end_time = my_keyboard.get_key()
  
       t1 = self.time()
  

  Not sure how to record correct responses in a smart way.. There was another version that had different number of stimuli: 1-back: 1 shape after the 'key' stimulus, 2-back: 2, etc... But I just don't think it's a very elegant solution

  The upper script is a good starting point for adding nice correct response logging
  You just have to define the correct response in the while loop,and compare it to the actual one. Again, here a way of doing it:

  
  # add before the while loop:
  trial_counter = 0
  stimuli = []
  
  
  # add in while loop
  trial_counter += 1
  stimuli.append(file)
  
  # define  correct response
  if stimuli[- (n+1)] == stimuli[-1] :
      #  n-1 defined by condition, which you should specify earlier
      correct_response = 'key'
  
  # check for given response
  if correct_response == rp_key:
       given_response = 1
  else:
       given_response = 0
  
  # for all relevant variables:
  exp.set('variable_name', variable)
  

  I removed feedback from this version, as I did not know how to bind it to the correct responses.

  Add another canvas, which only shows feedback text:

  # beginning of script:
  feedback_cv = canvas(exp)
  
  # in the end of while loop:
  if given_response:
      text = 'correct!'
  else:
      text = 'Incorrect!'
  
  feedback_cv.text('text')
  feedback_cv.show()
  self.sleep(300)
  

  I hope these pointers will suffice for you to get started. Btw. I am not sure how exactly the paradigm is supposed to work. I have to versions in my mind, that make both sort of sense, so in case you realize I explained the wrong one, sorry. But still, you should be able to modify the code for your needs.

  Good luck,

  Eduard

  
• eduard

  February 2015 edited 5:19PM

  I realized, it might have been better, if I just posted one coherent code snippet.
  Here you go!

  from openexp.canvas import canvas
  from openexp.keyboard import keyboard
  import random
  
  
  feedback_cv = canvas(exp)
  cv = canvas(exp)
  my_keyboard = keyboard(exp, keylist=['z', 'x'], timeout=3000)
  
  files_in_pool = ['star', 'circle', ...]
  trial_counter = 0
  stimuli = []
  
  # initialize timer
  t0, t1 = self.time(), self.time()
  while t1-t0 < 45000:
       trial_counter += 1
  
       # load image
       file = random.choice(files_in_pool)
       image_file = exp.get_file(file+'.png')
       stimuli.append(file)
  
       cv.image(image_file)
  
       # time response
       start_time = self.time() # as a reference for reaction time
       rp_key, end_time = my_keyboard.get_key()
  
       if stimuli[- (n+1)] == stimuli[-1] :
             #  n-1 defined by condition, which you should specify earlier
             correct_response = 'key'
  
       # check for given response
       if correct_response == rp_key:
             given_response = 1
       else:
             given_response = 0
  
  
       if given_response:
             text = 'correct!'
       else:
             text = 'Incorrect!'
  
       feedback_cv.text('text')
       feedback_cv.show()
       self.sleep(300)
  
      # for all relevant variables:
       exp.set('variable_name', variable)
  
       t1 = self.time()
  

  
• alebedev

  February 2015 edited February 2015

  Hi Dear All,

  Thanks a lot for all your help!
  After discussing the task with our group, we eventually decided to go for a pseudo-random sequence of events...
  The task is running relatively smoothly now I would say... However, I'm still struggling a bit with the logs (will manage it eventually, I hope).

  However, there is one particular problem that I still don't know how to fix...
  The thing is that I want to provide a feedback indicating which button (1-yes or 2-no) has been pressed without any correct/wrong indication [my experience suggests that without the feedback some subjects often get puzzled with this task]...

  The way it is supposed to look is as follows:
  

  However, when I start my paradigm I'm getting this error:
  

  Apologies for bothering... It's probably something straightforward (like incorrect assignment of the variables)...
  Could you please suggest what I have to change?

  I am attaching two archives:

  1.
  https://www.dropbox.com/s/zwk23k7rpw6jaek/nback_noFeedback.opensesame.tar.gz?dl=0
  [working version without the feedback]
  2.
  https://www.dropbox.com/s/43fmrsqcmbb246f/nback_wFeedback.opensesame.tar.gz?dl=0
  [an effort to implement this feedback with the error]

  Thank you very much for your time!

  Best,
  Alex
• alebedev

  February 2015 edited February 2015

  Ok..
  I think I found a way to implement the feedback eventually...
  Here is my working shortened version of n-back: https://www.dropbox.com/s/irvyfzub0j2wqzp/nback_wFeedbackDemoShortened.opensesame.tar.gz?dl=0

  However, I am not entirely sure how to fix the logs, which do not seem to generate the output I need...

  1. In particular, I'm not sure how to skip the first Ns and not to use them in accuracy calculation...

  2. Another issue is that the accuracy seems to be averaged across the trials, whereas I would like to have it calculated separately (e.g. for seq11, seq21, etc)...

  Right now the csv-output looks like this: https://www.dropbox.com/s/sndo6um9xzyiixp/subject-777.csv?dl=0

  I would really appreciate your suggestions Re these issues...

  Thanks a lot,

  Alex
• eduard

  March 2015 edited 5:19PM

  Hi Alex,

  In particular, I'm not sure how to skip the first Ns and not to use them in accuracy calculation...

  You can do it pretty much the same way, you decided which feedback to present when. That is, in the run_if line of your logger, put something like this:
  count_OneBack1 >= N.
  Opensesame keeps track how often an item is repeated. In your logfile, you will find these count-variables. Pick the one that suits you best.

  Another issue is that the accuracy seems to be averaged across the trials, whereas I would like to have it calculated separately (e.g. for seq11, seq21, etc)...

  The most obvious solution that comes to my mind is defining accuracy yourself. So, on every trial increment a counter of trials per sequence and of correct responses per sequence, calculate the average accuracy, save it as a new variable, e.g. avg_acc_block or something similar and finally log it.

  I suppose, you could also adapt the actual logger to make what you want it to, but I think this is more work than necessary.

  I hope this helped.

  Good luck,

  Eduard

  
• thpaim

  August 2017

  Hello Alex,

  Have you managed to finish your N-back test? I'm trying to implement a visual one but at the moment i'm strugling even to start. Do you still have your version? It would be of great help if could look at one implementation, but unfortunately your links have expired