Habit learning and conscious, declarative memory appear to run through largely separate systems in the brain — a person can build genuine skill and accuracy at a task through repeated cue-routine-reward exposure, entirely without being able to consciously recall ever having done the task before, or explain why a particular response feels right. This has a striking implication: habitual competence isn't necessarily built on top of conscious understanding, it can be built directly, in parallel, independent of it.
This also reveals something important about how fragile habitual learning is to context: it's tied specifically to the exact cue-and-reward conditions under which it was built, not to a general improvement in ability, which is why the same learned skill can completely fail to transfer the moment the surrounding format changes, even slightly.
To test whether Eugene Pauly, whose severe amnesia prevented him from forming new conscious memories, could still learn, Squire had him repeatedly choose between paired objects, one of which always concealed a "correct" sticker underneath. Eugene could never consciously recall having done the task before, or explain any reasoning behind his choices — but after weeks of repetition, his accuracy climbed to 95%, and he described his hand as simply "going for it" without knowing why. When the same objects were later presented all at once, outside their original paired format, Eugene was completely lost, proving the learning was tied specifically to the exact cue-and-routine context it was built in, not to a general improvement in memory or reasoning.