We report the results of an experiment probing the relative influence of centering and anchoring biases in reconstructive memory for line lengths. On each of 90 trials participants (N=120) viewed a target line, which they reproduced after a delay by adjusting a response line. We manipulated the starting size of this response line in three conditions: one provided an anchoring bias opposite the centering bias (expand condition), one in the same direction (contract condition) and one that provided no anchoring bias (control). We eliminated the centering bias in the expand condition, increased it in the contract condition, and showed an attenuated centering bias in the control condition. We discuss the implications for these results in relation to cognitive models of stimulus reproduction that employ the method of serial reproduction. We suggest that experiments of this type should carefully control for the possible influence of anchoring biases in reconstructive memory.