The age-by-complexity effect is the dominant empirical pattern in cognitive aging. The current report investigates whether a specific high-level mechanism---an age-related decrease in the reliability of episodic accumulators---can account for the age-by-complexity-effect, which is commonly assumed to be caused by an unspecific, low-level deficit. Groups of younger and older adults are compared in six reaction time experiments, using orthogonal manipulations of early cognitive difficulty (e.g., Stroop condition) and episodic demands (e.g., stimulus-response mapping). The predicted three-way interaction of age and the two factors was observed fairly consistently across experiments. A modified Brinley analysis shows that different regression slopes in old-young-space are required for conditions with low and high episodic difficulty. As a methodological contribution, a Brinley regression model following from certain simple processing assumptions is developed. It is shown that in contrast to a standard Brinley meta-analysis, the regression slopes in this model are not influenced by theoretically un-interesting between-experiment variance.