Hi All,
I need to calculate the working memory (WM) "threshold" capacity in terms of set-size (number of letters) eliciting 75% correct performance in a match task. I will assess this capacity on an individual basis (and, if possible, relatively quickly...) prior to a "main" experiment in order to use in the "main" experiment the set-size that elicits a 75%-performance level for each participant.
This problem resembles threshold estimate in psychophysics and I am aware of the existence of algorithms such as fixed-stimuli, adaptive staircases, adaptive MLE-based procedure (i.e. QUEST). However, I don't know of the use of such (or other) procedures in the domain of WM capacity.
Many WM papers just report "We used an adaptive staircase procedure to determine the WM set-size...". I would like to get acquainted with advantages/disadvantages of different procedures and algorithms in the WM-capacity domain. Does a "routine" procedure exist? Can anyone advice and/or point me to some relevant literature?
Thanks,
Francesco
See http://englelab.gatech.edu/
They have been doing some great work measuring WM.
Hi Francesco,
There are numerous ways to score wokring memory tasks if you adopt a procedure where you set a number of trials at various list lengths (e.g. three trials at each list length from two through six) and administer them all to each participant. Some relevant sources are Friedman & Miyake (2005; DOI: 10.3758/BF03192728) and St Clair-Thompson & Sykes (2010; DOI: 10.3758/BRM.42.4.969).
Some of the more continuous measures in these papers are perhaps best described as WM performance measures as opposed to strict capacity measures.
The adaptive staircase procedure you have seen in the literature likely refers to the dominant method for measuring WMC whereby participants are given a number of trials at the lowest list length, n, and are only given trials at list length n+1 if they achieved success on a majority of those trials. The value of n rises until a participant is unable to be successful at the pre-defined acceptance rate.
This method is very common within WM assessment so you have a good reason to administer a WM task in such a way. However, by administering all trials to all participants you have a number of options in how you measure performance and these alternative scoring options often have more desirable psychometric properties (see references above).
The primary advantages with adaptive-staircase terminating algorithm are that the time taken to complete the task can be minimised, and potential frustration from participants being given trials at list lengths they are unable to succeed at.
Hope this helps you in some way.
It Is fine to use the complex WM span task or WM updating task to measure WM capacity. By increasing the WM load continuously, you could determined the maximun memory setsize of each subject.
My concern is that WM performance could be improved with practice, so in the "main" experiment, the task performance might be better than you expected.
however, if your "main" experiment is not related to WM itself, it might not matters much to the results.
If you work with Matlab at all you may want to check out MemToolbox.
http://visionlab.github.io/MemToolbox/
Thanks you all for the suggestions! I'm diggin into the literature, which was suggested, and doing some exploration of the MemToolbox with Matlab.
- Badges
- Science method