cognitive fun!

3. On feedback. Feedback is very arguably a distractor in this test. However, it is also arguably a catalyst of learning. Suppose you are 49% confident you have an n-back, but decide, incorrectly, not to respond. Then suppose you are 51% confident, and decide, incorrectly, to respond. In a non-feedback version, you either realize your mistake after the session, in which significant time has elapsed, or you are kept oblivious. Compare this to a case where feedback is administered immediately, and the goal would be to enhance the 49% confidence and diminish the 51% confidence. With a loose behavioral interpretation, this hypothesis holds water.

2. On arrow keys. The principle difference here is that this version allows single-handed input, as opposed to two-hands. There will probably be effects due to the hands being controlled by different brain hemispheres, but this confound is overlooked. You are free to use two hands on the arrow keys as you wish, or use one hand, or foot, whichever (for what it's worth, the foot should be yet slower than the hand).

1. The presentation order in the original version is predetermined, and always has 20+n trials. This is likely necessitated by the need for rigorous comparison and analysis, hence minimizing the variation between confounds that may affect the independent variable (working memory assessment). As this version is not strictly monitored, the trial count is made flexible. Furthermore, as the original version had pre-determined quotas for targets, it is possible to predict the presentation order (for example, if two dual-modality targets have gone by, you are assured they will not appear again), decreasing cognitive load. A probabilistic approach mitigates predictability. Specifically, this n-back variant uses a 1/3 target probability, which means out of 20 trials, about 2 (1/9) are dual-modality, and about 4 (2/9) for each other modality. In the long run, the target/non-target ratios are similar to the original.