I've been reading on n-back's non-transferable working memory training, but if it's just used as a method of measuring the subject's working memory is it a good measurement? Personally, I have a poor working memory, easily forgetting what I was just speaking, how the arguments arrive at certain point, what I was just thinking that is important like a few seconds ago. Currently I've been playing n-back from time to time, and I am stuck on dual 3-back forever, never really able to pass through dual 4-back. Typically I don't have a habit of remembering stuff because I can always look up from search engines, and I take digital notes.
Yes. n-back performance correlates with short-term memory capacity as measured by standard tests. However, there is no standard implementation of the n-back task, which is why the literature is fairly inconsistent. Because it is not administered in a standardized fashion, it is not a good test that is comparable between studies. However, it is often used in working memory training.
Here is a paper that is critical of the n-back task. To be fair, although there is scientific consensus that there is insufficient evidence that "brain training" is effective, this is a publication from one of the more skeptical (which is a good thing) research groups. In this regard my example is somewhat biased, but I am open to discussing other perspectives. The group has also published some quite excellent criticism of how training experiments are not always well-designed, but I digress...
In cognitive training, it is important to distinguish between tasks that are optimized for testing and tasks that are optimized for training. A test task should optimally not vary in performance. The retest reliability should be high, while a training task should optimally improve performance over time and induce neuroplasticity in a consistent manner.
It is still open what makes a training task efficient. Some suggestions have been made, but criticized to an equal degree. One such concept is adaptive tasks. This describes when a task adapts the difficulty to the user depending on the user's previous performance. In summary, different n-back tasks vary in design and thus may vary in efficiency.
The problems of far and near transfer remain. Sometimes a battery of different tasks are administered simultaneously, but in such cases it becomes more difficult to attribute improvement on performance to specific tasks.