The aging of T cell memory is often considered in terms of senescence, a process viewed as decay and loss of memory T cells. How senescence would affect memory is a function of the initial structure of the memory repertoire and whether the clonotypes that make up the repertoire decay at random. We examine this issue using the T cell memory generated to the conserved influenza A epitope M1(58-66), which induces a strong, focused, but polyclonal CD8 T cell response in HLA-A2 individuals. We analyzed the CD8 T cell memory repertoires in eight healthy middle-aged and eight healthy older blood donors representing an average age difference of ∼ 27 y. Although the repertoires show broadly similar clonotype distributions, the number of observable clonotypes decreases significantly. This decrease disproportionally affects low-frequency clonotypes. Rank frequency analysis shows the same two-component clonotype distribution described earlier for these repertoires. The first component includes lower frequency clonotypes for which distribution can be described by a power law. The slope of this first component is significantly steeper in the older cohort. Generating a representative repertoire for each healthy cohort allowed agent-based modeling of the aging process. Interestingly, simple senescence of middle-aged repertoires is insufficient to describe the older clonotype distribution. Rather, a selective clonotype expansion must be included to achieve the best fit. We propose that responses to periodic virus exposure may drive such expansion, ensuring that the remaining clonotypes are optimized for continued protection.