Following on the extensive literature within experimental psychology on the reading of natural language texts, I have undertaken a series of experiments on the sight-reading of musical scores that have shown that spacing of information, the structuring of the musical discourse, and the predictability of design in a score can aid its legibility in a manner similar to what has been shown in the language domain. Cultural studies of reading —particularly the works of Saenger— point in the same direction; according to these, the change in Medieval textual scripts from scriptura continua at the beginning of the eight century to the adoption of canonical separations between words, phrases, or paragraphs (which had fully spread throughout Europe by the mid-fourteenth century) significantly decreased the cognitive load and time that had previously been needed to decode a script. Crucially, this eliminated the need for the ancient techniques of the praelectio (initial decoding of the text by reading it aloud) and rote repetition for its comprehension, triggering a whole new culture of private fluent reading. Equally, the literature on music sight-reading (although lacking in systematic research based on objective measurements of legibility of texts) has proposed, based on surveys and studies of expertise, a series of cognitive models of the activity that prime, as factors that distinguish proficient readers from beginners: the integration of discursive elements into higher-order meaning units, the ability to predict upcoming information, and the awareness of the structuring of the text. The experiments reported here compared readings using conventional scores with readings using novel scores where the suggested advantages of information separation, integration and predictability were implemented in the design. Fluency of performance was measured primarily in terms of numbers of mistakes, results showing that readers played more accurately with the novel scores. Other, more qualitative, measurements —such as spectrogram coding of tempo stability, blind expert judgment of performance quality, and participant self-assessments— all showed strong positive correlations with the measurements of numbers of mistakes, with the novel scores producing performances that were more fluent and ranked as more trustworthy and musically satisfactory by experts and readers alike. These results will still need to be extrapolated to many other musical practices, but they serve to open a debate on the conventions of music publishing as they stand, and are well placed to open new lines of research in score legibility and design.