Spatial numerical associations (SNAs) are ubiquitous. For symbolic stimuli, SNAs are shaped by cultural experience, including association with the Cartesian coordinate system. However, SNAs for numerosities are even found in infants and newborn chicks. Recently, the brain’s asymmetric frequency tuning hypothesis (BAFT, Felisatti et al., 2020) proposed that SNAs are really caused by a more fundamental perceptual property, spatial frequencies, for which many animals have hardwired low-level detectors in early vision, and whose processing is lateralized. This grounding in a basic perceptual quality is a valid numerosity signal because numerosity and spatial frequencies are typically confounded. Here we provide the first experimental evidence for the BAFT hypothesis by unconfounding spatial frequencies and numerosity. We compared stimuli that were equalized for their spatial frequency spectrum with standard numerosity stimuli. A go/no go task IAT task was used to avoid spatial response bias. Responses to small or large numerosities were paired with responses to arrow stimuli in mixed blocks, and mapping rules were switched between blocks, e.g., "respond only if left arrow or small numerosity". We observed a spatial-numerical congruency effect only for the standard stimuli. We conclude that spatial frequency information is a fundamental quality informing numerosity judgments.