Executing the homebound path is a major source of error in homing by path integration.

Path integration-the constant updating of position and orientation in an environment-is an important component of spatial navigation, however, its mechanisms are poorly understood. The aims of this study are (a) to test the encoding-error model of path integration, which focuses solely on encoding as a potential source of error, and (b) to develop a model of path integration that best predicts path integration errors. We tested the encoding-error model by independently measuring participants' encoding errors in distance and angle reproduction tasks, and then using those reproduction errors to predict individual participants' errors in a triangle completion task. We sampled the distribution of encoding errors using Monte Carlo methods to predict the homebound path, and then compared the predictions to observed triangle completion behavior. The correlation between predicted errors and actual errors in the triangle completion task was extremely weak, whereas an alternative model using execution error alone was sufficient to describe the observed errors. A model incorporating both encoding and execution errors best described the triangle completion errors. These results suggest that errors in executing the response may contribute more to overall errors in path integration than do encoding errors, challenging the assumption that errors reflect encoding alone. Errors in triangle completion might not arise from failing to know where you are, but from an inability to get back home. (PsycInfo Database Record (c) 2020 APA, all rights reserved).