1
$\begingroup$

I'm thinking of a study in which participants are randomly assigned to do task A then task B, or vice-versa.

I figured there will be a problem if doing Task A is really good practice for Task B, but doing task B isn't good practice for Task A. However, let's say I have good grounds for thinking Task A's practice effect on Task B will be the same as Task B's practice effect on Task A.

If I can assume that much am I clear to proceed? Or is there something else I should be worried about?

$\endgroup$
2
  • $\begingroup$ I'm a little unclear on what you're asking. Are you asking whether it's a problem if the practice effect of B on A is of a different magnitude than the practice effect of A on B? $\endgroup$ Commented Sep 30, 2013 at 15:45
  • $\begingroup$ If the practice effect of A on B is the same as the practice effect of B on A, and if I counterbalance between orders (say 30 participants doing AB, and 30 doing BA), is my methodology sound? Or is there some potential problem I haven't considered? $\endgroup$ Commented Sep 30, 2013 at 16:01

2 Answers 2

1
$\begingroup$

It really depends on what you mean with "take care".

If I reconstruct the situation you describe I assume the following:

  • There are two tasks A and B associated with one or several measurments of response quality that form a dependent variable of interest.
  • You assume that after finishing taks A, scores in task B will be elevated, but not the other way round. (This is a possible, but not a very common effect).
  • You randomly assign participants to order conditions, that is AB or BA.

If what you asusme is true, then you would find an interaction between task order and task type. There is nothing wrong about proceeding this way. The design does not eliminate practice effects, but it allows you to detect them.

Still, it would be informative to know more about what you are interested in. This might be a senisble or less sensible approach depending on what your research interests are.

$\endgroup$
1
  • $\begingroup$ It really was a hypothetical example. I accidentally obscured that by stating "I'm thinking of a study..." instead of "I'm imagining a study..." $\endgroup$ Commented Sep 30, 2013 at 15:59
0
$\begingroup$

I agree with @jank's answer. There's really nothing wrong with practice / order effects per se, and, in fact, the detection of these order effects can sometimes represent theoretically interesting contributions in their own right.

It is true that, without counterbalancing and when tasks A and B represent the within-subjects manipulation of interest, the order in which tasks A and B were presented becomes confounded with condition making inferences about the manipulation of interest ambiguous. However, in some cases, for example when tasks A and B are two measures and the primary focus is on a between-subjects manipulation, even this is not necessarily a concern.

If you're really concerned about order effects, the safest way to approach the administration of two tasks is to do it in entirely separate sessions. This is something my lab does routinely on tasks where we think taking one will influence future tasks. Of course, this sort of design is not always possible.

In short, what you do with respect to order effects really depends on the specifics of the study design. Having order effects of unequal magnitude should not be a concern in the slightest, especially if you have designed your study in such a way as to allow you to measure their magnitude.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.