The psychological functions assessed by substitution tests, and the age-related performance decline, are not well understood. Here several aspects of long-term memory were manipulated across younger and older adults. A 45-page Digit-Symbol test was employed. Each page contained a 9-item digit symbol code-table and 9 response items. There were 9 study conditions with each condition deployed across 5 pages, or trials, of the test. The conditions were formed by crossing two within-subjects factors, each with 3 levels. The first factor, Digit Order, pertained to having the code table digits in numerical order vs. a pseudo-random order fixed across trials vs. a pseudo-random order that varied across trials. The second factor, Symbol Pairing, pertained to having a fixed digit-symbol pairing across trials vs. having a varying digit-symbol pairing across trials vs. having a novel set of 9 symbols introduced on each of the 5 trials. Including the additional factor, Age, resulted in a 2 x 3 x 3 mixed randomised block design. The older group was slowed, F(1, 22) = 17.267, p < .001, and overall-performance was poorer when the digits were arranged non-numerically, F(1,44) = 55.403, p < .001. An Age by Symbol-Order interaction indicated that use of novel symbols disadvantaged only the older participants, F(1, 44) = 6.577, p = .014. While there was no evidence that incidental paired-associate learning or spatial memory affect digit-symbol performance, symbol familiarity may be important to digit symbol test completion in older adults. The benefit of ordinally arranged digits in the coding table highlights a fundamental process difference between Digit-Symbol and Symbol-Digit test formats.