Examining the Robustness and Generalizability of the Shape Bias: A Meta-Analysis

Abstract

The “shape bias” – the tendency to generalize new nouns by their shape rather than features like colour or texture – is considered a cornerstone of early noun learning. However, the magnitude, development, and cross-cultural generality of this bias remain debated. We conducted a pre-registered meta-analysis and meta-regression of 71 studies to synthesize the evidence base. Our analysis revealed a robust overall effect, confirming the existence of a shape bias. However, this effect was characterized by extreme between-study heterogeneity ( I ²  = 0.95). Critically, meta-regression analyses failed to support hypothesized developmental changes, with a constant model across ages providing the most parsimonious fit. The literature is overwhelmingly dominated by studies of English-speaking children, preventing meaningful tests of cross-linguistic or cross-cultural variation. Diagnostic tests were also consistent with the possibility of publication bias, though these patterns can also arise from methodological heterogeneity and should be interpreted with caution. Together, these findings indicate that procedural variation, and potentially publication bias, more than currently testable theoretical moderators, account for the variability in the literature, severely limiting the ability to distinguish between theories of the shape bias’s origins.