A perennial preoccupation in the study of early writing systems is the degree to which these earliest forms of written communication reflect speech. Nowhere is this concern more keenly present than in study of the earliest writing from Mesopotamia, known as proto-cuneiform, for which the vast gulf that separates speech from writing raises questions about the very language that underlies the script. Overlooked in the debate over the presumed Sumerian basis of proto-cuneiform is the more fundamental question of how the first scribes conceived of written language in the first place. Was writing, in its original conception, language-based? Or was it a system in which graphs first and foremost represented things and only secondarily the words attached to those things? That is, what was the nature of the original representational relationship between symbol and referent. This paper explores these questions and addresses the primary evidence for these two propositions as well as some underlying theoretical considerations and cross-cultural parallels for analogous semiotic systems. The paper will also explore the philological speculations and experiments that the first scribes engaged in as they attempted to apply the incipient writing system to the expression of new genres, including literature.