Abstract: Recently, the development of human numerical ability has attracted more and more attention, and a great deal of advance of number processing-specific phenomena has been reported, one of which is the Spatial-Numerical Association of Response Codes effect (SNARC effect). Instinctively, number processing is a numerical symbolic operation without additional information. But, in fact, a close relationship between number and space has been found, which demonstrates a dependency of number processing on the spatial character of objects. Some researchers explained that numbers are represented along a horizontally “mental number line”, on which small numbers are on the left of the line, and large numbers are on the right. However, since the SNARC effect was also found on the vertical direction, someone proposed that the spatial representation of numbers is a number map rather than a number line. Furthermore, the SNARC effect was likewise discovered when doing the shallow processing of number (Fischer, 2001; Shen et al., 2006). Besides, previous studies (e.g., Wood & Willmes, 2005) have found the SNARC effect in many other numerical symbols, such as auditory number words, visual Arabic numeral, and visual number words. However, the SNARC effect of non-symbolic number which is one of the numerical formats has not been studied so far. Although researchers have paid attention to the relationship between the eye movement and cognition in recent 20 years, we still need more eye movement researches in number processing. The present study explored the influence of processing levels on the SNARC effect of dot matrixes by the technique of eye movement. The present study includes two within-subjects design experiments, to study the SNARC effect on the horizontal and vertical directions, respectively. Experiment 1 investigated the SNARC effect of non-symbolic number, in which participants were asked to response to the detective stimuli (on the left or right) following the irrelevant stimuli (dot matrix) with two different processing levels (deep processing condition and shallow processing condition). The eye movement response time was recorded respectively for deep processing (the comparison task) and shallow processing (the dice scanning task) to examine whether the SNARC effect appeared or not. The results showed that SNARC effects appeared in both conditions of horizontal dimension, further analysis indicated that the strength of SNARC effects from both processing levels was similar. In Experiment 2, the detective stimuli were represented in vertical dimension, and the tasks were the same to those in experiment 1. We recorded eye movement time for both upward and downward responses. The result showed us that there was no SNARC effect in both conditions. In conclusion, the present study suggests that the SNARC effect of dot matrix was weaker than that of Arabic number, and the processing level almost had no impact on the SNARC of dot number dueling to the interaction of the characteristic of the dice and the direction, while the role of dimension was crucial.

Key words: numerical cognition, processing level, dot matrix, eye movement, SNARC effect