Though it seems absolutely clear what randomness is, there is no unique understanding of it. Even mathematics knows several different concepts.
"Every finite symbol string is random under some nontrivial definitions of randomness. Every finite symbol string is nonrandom under some nontrivial definitions of randomness."
These two paradoxical conclusions in Gilmore (1989, p. 339) underline the fact, that it might be worth the effort to consider the reasons of randomization and the principles of the process. It is the purpose of RQube to provide several criteria to judge the quality of a random stimulus series. According to these criteria RQube generates as many series as you want. These may easily be implemented in your experiment control program.
Randomization is a basic procedure in experimental design. Considered the importance of it, it is odd to say, that -- except for a few articles in parapsychological research -- the topic did not receive very much attention. RQube tries to fill that gap. Not only by providing a practical tool, but also by reflecting about the reasons and demands of different approachesof randomization.
Why different ways of randomization?
If subjects are able to predict an upcoming event, they may begin to prepare their actions in advance. As a result, the stimulus may not continue to elicit the process under investigation any more. Anything that would have been interesting takes place before a stimulus presentation.
Avoid sequence effects
In many cases the sequence of two treatments A -> B has not the same consequences than B -> A.
In order to prevent habituation, a stimulus must not occur often and there have to be frequent lternations between the stimuli.
Avoid distribution effects
The distribution of conditions is not homogenous across the total duration of the experiment. A consequence may e.g. be that some conditions are presented mostly at the end of the experiment when subjects become tired.
Why a unique series of conditions for each subject?
Any trial series may be skewed. You may rule out many problems by using RQube. But you can only rule out problems that you anticipate. The more different trial series you use, the less likely your data is influenced by a biased randomization.