Causal reasoning is a principal higher-cognitive ability of humans, however, much remains unknown, including (a) the type (systematic versus intermixed) and order (inductive-thendeductive or vice versa) of experience that best achieves causal-chain extraction; (b) how inferences generalize to novel problems, especially with one-shot experience; and (c) how metacognition, reflected in uncertainty of one's knowledge, relates to actual knowledge. We tested people on a realistic cancer biology task (e.g., 'seroc' chemicals inducing tumors with subsequent effects). Systematic experience was superior, with some evidence that the inductive-then-deductive order promoted stronger one-shot generalization. Notably, uncertainty was decoupled from actual knowledge, with the deductive-then-inductive group being overconfident, likely reflecting lack of awareness of the inductive component; while those with successful one-shot generalization held lower confidence, reflecting generalization with minimal experience, while remaining skeptical. Our findings clarify processes underlying causal reasoning, and reveal a complex relationship between causal reasoning and metacognitive awareness of it.