Shape of SNARC: How Task-Dependent Are Spatial-Numerical Associations? A Highly Powered Online Experiment

Spatial-Numerical Associations are fundamental to numerical cognition. The Spatial-Numerical Association of Response Codes (SNARC) effect denotes faster responses to small/large numbers with the left/right, respectively (Dehaene et al., 1993). It is typically measured in magnitude classification (MC: smaller vs. larger than 5) or parity judgment (PJ: odd vs. even) with numbers 1–9 (excluding 5). These tasks differ in magnitude and parity processing, compatibility effects, and other aspects. In this highly powered online experiment, data from 611 participants in a within-subjects design demonstrated a rather categorical MC-SNARC (i.e., similar left-hand advantage for 1–4, and similar right-hand advantage for 6–9) and a continuous PJ-SNARC (i.e., right-hand advantage increasing with magnitude). Accordingly, the PJ-SNARC can be modeled using linear regression, while the MC-SNARC requires a categorical or mixed analysis. Explained variance and split-half reliability were higher for the MC-SNARC than for the PJ-SNARC. We replicated the Numerical Size Effect (slower responses for larger numbers; Brysbaert, 1995) in MC and PJ, the Numerical Distance Effect (faster responses for larger numerical distance between stimulus and reference; Moyer & Landauer, 1967) in MC, and the Odd effect in PJ (faster responses to even numbers; Hines, 1990; potentially masked by the Numerical Distance Effect in MC), but not the MARC effect (faster responses to odd/even with the left/right, respectively; Nuerk et al., 2004). The MC-SNARC and PJ-SNARC correlated slightly positively (r = 0.17), suggesting partly common underlying magnitude representations. Altogether, the SNARC effect appears highly task-dependent and should be modeled differently depending on the task.