This study examined 25 university students' use of addition to solve large single-digit subtractions by contrasting performance in the standard subtraction format (12 - 9 = .) and in the addition format (9 + . = 12). In particular, we investigated the effect of the relative size of the subtrahend on performance in both formats. We found a significant interaction between format, the magnitude of the subtrahend (S) compared to the difference (D) (S > D vs. S < D), and the numerical distance between subtrahend and difference. When the subtrahend was larger than the difference and S and D were far from each other (e.g., 12 - 9 = .), problems were solved faster in the addition than in the subtraction format; when the subtrahend was smaller than the difference and S and D were far from each other (e.g., 12 - 3 = .), problems were solved faster in the subtraction than in the addition format. However, when the subtrahend and the difference were close to each other (e.g., 13 - 7 = .), there were no significant reaction time differences between both formats. These results suggest that adults do not rely exclusively and routinely on addition to solve large single-digit subtractions, but select either addition-based or subtraction-based strategies depending on the relative size of the subtrahend. (C) 2009 Elsevier B.V. All rights reserved.