Obviously in the limit every test shares something with some other test. But it's not like it's hard to find cases where practice on one task fails to generaliz

Obviously in the limit every test shares something with some other test. But it's not like it's hard to find cases where practice on one task fails to generalize to another task: in fact it's the norm, and it's exactly this empirically observed lack of generalization that motivated the concept of IQ and g in the first place. For example, see https://www.ncbi.nlm.nih.gov/pubmed/27474138 https://www.ncbi.nlm.nih.gov/pubmed/27474138