<p>Geometric morphometrics is widely used in mammalogy, taxonomy and other biological disciplines to compare organismal forms. For accurate shape analysis, however, measurement error (ME) must be carefully assessed. To this aim, there are different approaches, including a recently proposed method that partitions systematic and random ME and tests their variance ratio to detect a significant directional error (i.e., a bias). Using 2D data on ventral views of adult marmot crania with multiple replicate landmark digitizations, I explored the effect of time lags between digitizations and assessed the impact of potential biases due to digitization error. I found a statistically significant bias, whose magnitude showed a non-linear increase with the time interval between two digitizations. This observation supports recent evidence for the potential occurrence of a ‘visiting scientist effect’, when landmarks are digitized in separate visits to museum collections. However, the bias impact depends on the magnitude of the biological variation being investigated: thus, while a small systematic ME may be enough to overturn conclusions on sexual dimorphism, as demonstrated in previous research on a subsample of the current dataset, a similar significant bias has a mostly negligible effect on tests of interspecific mean shape differences, which in marmots are much larger than sex differences. This suggests that, while statistical significance in the test for systematic ME is a useful warning, researchers might also evaluate the ratio between the effect size of the biological signal of interest and the magnitude of the bias to accurately interpret the impact of ME.</p>