Abstract
To investigate effects of surface notation on basic numerical skills we examined number naming, magnitude selection, and simple arithmetic performed by adult Chinese-English bilinguals born and educated in China. Stimuli were presented using either arabic (7 + 8) or “mandarin” (㈦ + ㈧)symbols and participants were cued to respond either in English or Chinese. The naming task demonstrated that the mandarin characters were as easy to identify as the arabic digits, but for both arithmetic and magnitude selection there were faster RTs and fewer errors overall with arabic notation. Arabic notation also produced smaller problem-size effects in arithmetic and a smaller split effect in magnitude selection relative to mandarin notation. These results suggest that retrieval processes in the arithmetic and selection tasks were more efficient with arabic than mandarin stimuli. Arithmetic RTs were substantially slower with English than Chinese responses given either arabic or mandarin stimuli, but the English-language cost was greater with mandarin stimuli. The form of the Notation Language RT interaction is consistent with language-specific Chinese and English number-fact stores (“arithmecons”) that were differentially accessible as a function of notation. Naming RTs also presented a significant Notation Language interaction due mainly to slow RTs to produce English number names for mandarin stimuli. These Notation Language interactions are not easily reconciled with the standard version of McCloskey's (1992) model of number processing, which holds that numeral reading and arithmetic performance are based on a single, abstract-semantic code regardless of input or output conditions. Instead, the results suggest that different Notation Language combinations were mediated by independent associative paths that varied in strength and efficiency as a function of prior experience.