Discussing the organization of developmental switches, West-Eberhard (2003) proposed the use of regression residual plots to locate the neutral point of the switch, which is characterized by maximum genotypic influence on the resulting phenotype. However, statistical artifacts due to measurement error in nonlinear models might account for a substantial proportion of the increase of residuals variance at the switch point. Simulations based on field data show that increases in residual variance occur as artifacts when normal amounts of measurement error are present, even in absence of any genotypic variance. The results suggest that interpretation of nonlinear variation in thresold traits is problematic and requires considering this statistical effect. A method to estimate the weight of genotypic contribution to residual variance is proposed, and its assumptions and limitations are discussed.