The inability to develop, strengthen, and access associations in memory that allow for the rapid and accurate retrieval of answers to basic addition problems is a distinguishing characteristic of a mathematics learning difficulty. The 'two-factor theory of math fact learning' (Robinson, Menchetti, & Torgesen, 2002) proposes that a weakness in semantic or phonological processing relating to number underlies such difficulty. The empirical support for this theory has been limited. In this study the basic addition performance of five adolescent students still reliant on counting was examined. A regression analysis of reaction times to counting trials revealed counting-speed to be an important factor in helping to explain why practice had not led to retrieval. The findings are discussed in terms of advancing the two-factor theory of math fact learning and implications for instruction are considered.