The catch-all underestimation bias: 'availability' vs. 'category redefinition' hypotheses