Why far transfer is harder than it looks
Practice gains are not the same as transfer
When a cognitive training programme reports that users improved by 20 percent, the first question is: improved on what? In many studies, the measure is close to the task users trained on. That can be useful, but it is not the same as evidence for broad cognitive change.
Repeated exposure produces procedural learning. Users learn the interface, timing, distractors and local strategy. Those gains are real, but they can remain tied to the original format.
What far transfer asks
Far transfer asks whether improvement carries into tasks that were not part of training, with different formats, demands or contexts. That is the question people usually mean when they ask whether brain training works.
Near-transfer and far-transfer trends can both look upward on simple charts. They need to be separated with changed formats, changed rules and delayed checks.
How transfer should be checked
IQ Mindware uses game swaps, boundary probes and delayed re-checks. The goal is to test whether performance survives condition changes, not only whether in-task scores rise.
A gain that does not survive condition changes is familiarity evidence, not far-transfer evidence.
Delayed checks matter because immediate improvement can reflect short-term tuning to the task. A more useful signal is performance that remains available after time and format changes.
The claims boundary
Transfer remains an active research area. The evidence base is mixed and should be interpreted with caution. IQ Mindware uses conservative language aligned to published boundaries and its own proof posture.
Protocol logic is documented at /proof/#protocols, and claims boundaries are explained at /proof/#claims.