Stroop-like paradigm was applied. Sixteen undergraduate students participated in Experiment 1. They were asked to compare the physical size of two two-digit numbers, as well as the numerical magnitude of the two numbers （e.g., 23 vs. 78）. Four hundred two-digit pairs consisting of the numbers between 21 and 98 were presented in Arabic notation, excluding numbers that contained 0 as unit （e.g., 30, 40） or the same digits for decade and unit （e.g., 22, 33）. We manipulated the ratio of the physical size of two numbers （i.e., 1:1.1 vs. 1:1.2）, the congruency of magnitude size and physical size, and the unit-decade-compatibility. To avoid confounding of the distance and problem size effect, we controlled overall distance and problem size between compatible and incompatible pairs. In addition, we also conducted Experiment 2 to avoid interference of intentional and unintentional processing. Twenty undergraduate students were asked to compare only the physical size of two two-digit numbers. Experimental design and procedure were same as Experiment 1, except the ratio of physical size of two numbers had three levels containing 1:1.1, 1:1.25 and 1:1.5.

The results from the magnitude comparison task in Experiment 1 replicated those found in previous studies involving the same tasks. Tasks involving physical size comparison in both Experiment 1 and Experiment 2 showed the effect of magnitude and physical size congruency. The effect supported holistic processing. The unit-decade-compatibility effect was evident for congruent number pairs, but there was no or a reverse effect for incongruent number pairs.

In sum, two-digit number processing can be componential as well as holistic in both unintentional and intentional processing of numerical magnitude.