In this thesis, we defend the notion that cognitive resources available for strategy execution do not only depend on the participant and current task demands but also on prior task demands. Strategy performance should thus be less efficient when the previous strategy was difficult (i.e., sequential difficulty effects). The notion of sequential difficulty effects was tested in four experiments with computational estimation (i.e., estimating the solution to arithmetic problems by rounding the operands). We found that execution of a mixed-rounding strategy on two-digit addition problems (i.e., rounding one operand down and one operand up) was less efficient after an easy rounding-down strategy than after a difficult rounding-up strategy. Moreover, the effect was stronger in individuals with less efficient working-memory capacity and in Alzheimer patients. These results confirm the existence of strategy sequential difficulty effects and suggest that working memory is involved.