Domestic dogs show flexible left-to-right spatial-numerical association

Problem
Spatial-Numerical Association (SNA), the tendency to associate smaller numerosities with the left side of space and larger ones with the right (often termed the "Mental Number Line"), has been documented across phylogenetically distant species, including birds, fish, insects, and primates. While evidence suggests this bias may be an evolutionarily ancient, non-symbolic mechanism independent of language and education, research on SNA in domesticated mammals remains sparse. Previous studies in pigs yielded negative results, potentially due to reliance on spatial rather than numerical cues. Furthermore, while SNA has been observed in newborns and animals with limited experience, it remains unclear whether these associations persist or are reshaped in adult domesticated mammals with heterogeneous individual experiences and prolonged human interaction. Additionally, the functional utility of SNA—specifically whether numerical spatialization confers an advantage in quantity discrimination efficiency—has not been directly tested in non-human animals.

Methodology
The study investigated SNA in 49 adult pet dogs (Canis familiaris) using a spatial-numerical congruency paradigm adapted from previous work in chicks, honeybees, and zebrafish. The research was divided into three experiments:

• Experiment 1 (SNA Assessment): 24 dogs were trained to locate a food reward behind a central panel displaying a reference numerosity of 8 dots. In unrewarded test trials, dogs were presented with two identical lateral panels displaying either a smaller numerosity (4 dots) or a larger numerosity (16 dots). To control for side biases, only dogs that successfully circumnavigated the central panel from both left and right during training were included. Dogs were classified as "SNA-consistent" only if they chose the left panel in the "smaller-than" condition and the right panel in the "larger-than" condition.
• Experiment 2 (Numerical Discrimination): 14 dogs from the first experiment underwent a simultaneous numerical comparison task to assess quantity discrimination abilities. They were tested on discriminations with a ratio of 0.50 (e.g., 4 vs. 8) and a more difficult ratio of 0.73–0.75 (e.g., 6 vs. 8). The study analyzed whether performance in the SNA task correlated with performance in these discrimination tasks.
• Experiment 3 (Control for Continuous Variables): 25 dogs were tested using the same paradigm as Experiment 1, but with stimuli designed to control for continuous visual variables. Density and total perimeter were equated across numerosities, while cumulative area and element size varied inversely with numerosity (e.g., the smaller numerosity consisted of larger elements with greater total area). This ensured that any spatial bias could not be attributed to physical extent or element size.

Data were analyzed using binomial tests against chance levels and Generalized Estimating Equation (GEE) models to account for repeated measures and assess the influence of attention and experimental conditions.

Key Results

• SNA Presence: In Experiment 1, the number of dogs exhibiting the SNA-consistent pattern (left choice for smaller, right choice for larger) significantly exceeded chance expectations (11/24, p = 0.042).
• Independence from Continuous Variables: In Experiment 3, the SNA-consistent pattern persisted even when density and total perimeter were controlled (12/25, p = 0.022). The results ruled out explanations based on cumulative area or element size, as these physical variables would have predicted a pattern opposite to the numerical association.
• Correlation with Discrimination Ability: In Experiment 2, dogs performed above chance in the easier 0.50 ratio discrimination but not in the more difficult 0.73–0.75 ratio. Crucially, a GEE model revealed that dogs showing a consistent SNA in Experiment 1 significantly outperformed those that did not in the difficult 0.7 ratio discrimination (Odds ratio = 9.696, p = 0.026). No such correlation was found for the easier 0.5 ratio.
• Lateralization Control: The study confirmed that the observed SNA was not a result of fixed individual side biases, as the classification required flexible, context-dependent responses (choosing left for small, right for large).

Significance and Claims
The paper claims to demonstrate that adult domestic dogs possess a flexible, left-to-right Spatial-Numerical Association, extending the evidence of this cognitive trait to domesticated mammals. The findings suggest that numerical spatialization persists despite extensive individual learning and interaction with humans, supporting the hypothesis that SNA is a widespread property of brains endowed with a "Number Sense."

Furthermore, the study provides preliminary evidence that numerical spatialization is functionally linked to the efficiency of magnitude processing. The correlation between SNA consistency and performance in difficult numerical discriminations suggests that organizing numerical magnitudes along a spatial continuum may facilitate quantity processing, particularly when the task approaches the discrimination threshold. The authors note that these results support the Right Hemisphere Dominance Model of SNA, which posits that the right hemisphere's specialization in both spatial and numerical processing underlies this association. The authors maintain a modest tone regarding the functional link, acknowledging the small sample size for the correlation analysis and calling for future studies with larger samples to directly test whether SNA facilitates quantity discrimination.