Background: Handwriting legibility systematically biases evaluations in that highly legible handwriting results in more positive evaluations than less legible handwriting. Because performance assessments in educational contexts are not only based on computerized or multiple choice tests but often include the evaluation of handwritten work samples, understanding the causes of this bias is critical. Aims: This research was designed to replicate and extend the legibility bias in two tightly controlled experiments and to explore whether gender-based inferences contribute to its occurrence. Sample(s): A total of 132 students from a German university participated in one pre-test and two independent experiments. Method: Participants were asked to read and evaluate several handwritten essays varying in content quality. Each essay was presented to some participants in highly legible handwriting and to other participants in less legible handwriting. In addition, the assignment of legibility to participant group was reversed from essay to essay, resulting in a mixed-factor design. Results: The legibility bias was replicated in both experiments. Results suggest that gender-based inferences do not account for its occurrence. Rather it appears that fluency from legibility exerts a biasing impact on evaluations of content and author abilities. Conclusions: The legibility bias was shown to be genuine and strong. By refuting a series of alternative explanations, this research contributes to a better understanding of what underlies the legibility bias. The present research may inform those who grade on what to focus and thus help to better allocate cognitive resources when trying to reduce this important source of error. (Contains 1 table, 2 figures and 5 footnotes.)