Yanfei Yu¹ (), Urvi Maheshwari², David Barner³; ¹University of California, San Diego, ²University of California, San Diego, ³University of California, San Diego

Humans can represent large quantities imprecisely (Cordes et al., 2001; Gibbon, 1977), but it is debated whether this reflects a single amodal system or multiple modality-specific capacities. The Approximate Number System (ANS) account proposes a modality-independent number sense, with similar acuity across visual, auditory, and tactile formats (Barth et al., 2003; Tokita & Ishiguchi, 2016). In contrast, the Sensory Integration System (SIS) theory argues that numerical estimates arise indirectly from modality-specific computations of correlated magnitudes, such as area or duration (Gebuis et al., 2016). The present study tested these views by asking whether feedback-based calibration of number estimation in one modality (vision or audition) transfers to the other. We employed a 2 (Inducer: overestimated vs. underestimated) × 2 (Modality: visual vs. auditory) × 2 (Session: pre- vs. post-calibration) mixed design, with inducer as a between-subjects factor. Adults (N = 61 and 60 in Experiment 1 and 2) estimated quantities of dots (10–97) and tones (10–37) before and after calibration. During calibration, participants were told that the inducer displays contained 42-48 dots in Experiment 1 and 20-28 tones in Experiment 2, although the actual stimuli contained either 3/4 this N (overestimated inducer) or 4/3 this N (underestimated inducer). Experiment 1 showed increased visual and auditory estimates after calibration in both inducer conditions, consistent with a general priming effect of large digits. However, only visual estimation showed an inducer x session interaction, suggesting a within-modality calibration effect that did not transfer to audition. In Experiment 2, auditory estimates increased after calibration only in the overestimated condition, and we observed transfer to visual estimation for values near the inducer set size, compatible with a modality-independent number representation argued by ANS. Ongoing work is exploring the possibility that the strong priming effect in Experiment 1 may have masked cross-modal calibration, producing weaker transfer.