Many of the tests used to assess executive functions are prone to training effects. In conducting a case-control study of the impact of cognitive training on executive functions, is it important to adjust data for training effects? If so, what are the best methods to use to accomplish such adjustments.
Firstly, in the case of cognitive training, you'd expect training effects to alter and improve performance on tests of executive functioning. So why control it there?
Secondly, if you're planning to do a before-after type of study where training effects need to be controlled and not anticipated, then I'd suggest looking at the specific tests you'd want to administer. For instance, if you're using the Iowa Gambling Task, you can easily change the penalty decks over two versions of the same experiment. Training effects should get cancelled out by that scheme. But if you plan to use something like the Trail Making test, then I'd suggest using a parallel version so that training effects do not mar your results. Hope this helps : )
Hansika Kapoor:
You have made some very insightful comments.
Thank you so much for your response.
Kerry Lee
Thank you for your very insightful answer. It is unlikely I will have a neuroimaging component in my study. I think neuroimaging would have been a great complement.
This is a great question! Any psychological test is going to produce a practice-related effect, as you know. The best way to control for this within the context of cognitive training research is to have a good control group to compare against your tx group -- thereby cancelling out the practice-related effects.
In addition, and an equally important point is that you have to have a research design that allows to assess for any transfer effects produced by your tx. That is, performance gains may be divided into near transfer (i.e., improvement on other WM tasks, such as a performance increase on a verbal WM memory task translating into improved performance on an untrained spatial WM task) and far transfer (i.e., improvements in other domains that overlap with but are separable from WM, such as fluid intelligence). If you see that your tx produces effects on a variety of tasks that measure the same construct then you can be more confident that you made a real change that is not specific to a task.
Hope this helps,
Michael