A study of processing load in single-digit number comparison

Lepoittevin, Samuel [UCL] Masson, Nicolas [UCL] Andres, Michael [UCL] et al.

In the digit comparison task, it is a robust finding that for a fixed numerical distance between the two to-be-compared numerals, response latencies increase regularly with their value. This “magnitude” or “size” effect has notably been interpreted as reflecting a reduction of the discriminability between the magnitudes conveyed by the digits, as their numerical value increase. But so far, this variability in the difficulty of digit comparison has only been studied using mental chronometry. However, reaction times, as an output performance, constitute quite an indirect measure of this difficulty. The present study thus aimed to use a more direct approach - namely: the psychosensory pupil response - to track the cognitive cost associated with Arabic digit comparison. To do so, we asked 20 participants to perform an Arabic number comparison task, in which the distance effect was controlled by keeping fixed the units between the two numbers to be compared (e.g., 1 vs. 3, or 5 vs. 7). The numbers ranged from 1 to 13. The participants had to respond verbally on which side was the largest - or the smallest - as fast and as accurately as possible. We collected response latencies and recorded the pupil size variations. In the analyses, we focused on the trials containing only single-digit numerals. First, as predicted, our results clearly showed that the number magnitude influenced both the response latency and the pupillary response, confirming that the difficulty of comparing digits increases with the magnitude they represent. Secondly, we found that odd-numbered pair comparisons were associated with greater pupil dilation, suggesting that comparing odd digits would be more cognitively demanding than comparing even ones. But intriguingly, we didn’t find any evidence of an impact of this parity effect on the response latencies. We conclude that pupillometry (1) contributed to tracking the cognitive cost associated with digit comparison and confirms that this cost increases with the magnitude they convey. Moreover, pupillometry (2) also revealed a novel effect, suggesting that a parity-related dimension would also impact the cognitive cost of digit comparison. Future research should investigate whether - and in which condition - this parity dimension would impact the reaction times. This could allow us to refine the mechanistic hypotheses about the processes underlying the comparison of symbolic numbers.