Some of this depends on who you believe regarding reading practices (and when/where you're looking). To paraphrase/copy some of notes and writing I have on the topic (mostly in the Middle Ages):

Paul Saenger has written about this topic quite a bit ("Silent Reading: Its Impact on Late Medieval Script and Society," Viator 13 (1982), 370-377; Paul Saenger, Space Between Words: The Origins of Silent Reading, Stanford 1997). In his view, from antiquity to the early Middle Ages the primary mode of reading was oral recitation. Part of this claim is based on how ancient writings often lacked clearly defined spacing between words—this makes silent reading difficult, but actually isn't too bad if you're reading by sounding letters aloud. Ivan Illich, In the Vineyard of the Text (Chicago: The University of Chicago Press, 1993), in talking about not just reading but writing in the Middle Ages, claims that much of the latter would also be done orally; “Writing—like reading—remained a mumbling activity.” (90)

Textual sources suggest that as late as the twelfth and early thirteenth centuries the written word was typically intended to be read aloud. This affected all areas of texts' production and use, from the ways in which words were written to the organization and lending practices of libraries (oral reading, being slower than silent reading, required far longer loan periods).

Again, in Saenger's view it was only by the late thirteenth and early fourteenth centuries that the practice of silent reading began to overtake oral recitation as the standard means of using a text. This method of silent reading then alters the shape and nature of texts (as the texts themselves altered the nature of reading). More legible scripts, clearly defined word spacing, and libraries that catered to a silent environment (with shorter loan periods) were just some of the substantial changes that resulted from this new form of interacting with the written word.

BUT—I know there are critics of Saenger's claims, who suggest that the notion of reading out loud has been overstated.