Higounet (2003) states that the writing is at the foundation of the human sciences. For a discipline, in particular, that integrates the studies of language, the Historical Linguistics, the writing is the cornerstone. Without the written records that have survived throughout history, there would be no diachrony of language. Nevertheless, very little effort among the diachronic studies in the lusophone domain has been dispensed to the writing as an autonomous linguistic object and not as an instrument for another purpose (historiographic or phonetic, for example), perhaps reflecting the idea found in some classical linguistic works according to which there was a supposed (ortho)graphic chaos in the Portuguese of past synchronies. This work intends, therefore, to contribute with this field through a graphemic analysis. In order to do so, however, it was first necessary to go through written language in two ways: reflection and analysis. As far as reflection is concerned, a critical review of the treatment of written language in linguistic studies is undertaken, and the main ideas focused on in this study about the functioning of writing are examined, especially as regards its mechanism in the history of language. The authors responsible for these main ideas to think a theory that allows to understand, with the greatest precision, its operation, mainly in the diachronic axis, are notably: Vachek (1989b, 1989c, 1989d) for a linguistic theory of writing; Cagliari (1996, 2001, 2001b, 2001c, 2001d, 2004, 2015) for a theory of orthography; Frago Gracia (2002), Sánchez-Prieto Borja (1998, 1998b, 2008) and Ramírez Luengo (2012b, 2013, 2014, 2015, 2015b) for a theory of (ortho)graphy concentrated on Historical Linguistics. As for the second way, constitution, a diachronic corpus is revised and re-edited, the official letters of Rio Grande do Norte (Morais de Melo, 2012), generating a set of approximately 26,000 words in 129 letters, from which 44.18% are unpublished. All the documents of this re-edition were written in Rio Grande do Norte between 1713 and 1950 and are presented in facsimile and semidiplomatic edition with a justalinear lesson. Once these two steps are fulfilled, the analysis is executed: a study of the use of alphabetic graphemes without phonemic transcendence is carried out on those words for every quarter of a century during the 250 years that the letters cross, and according to 23 fixed patterns of analysis, like the pattern {V<N>C-[<h>/(<s>#)]} to control the use of <m> or <n> before consonants, and the pattern {C/V<ãe>#/<s>} for the examination of the graphemic productions of value /ãj/. A software developed for this research ran the 23 patterns over the approximately 26,000 words in order to verify the existence of graphical tendencies or, on the contrary, to confirm the supposed state of (ortho)graphic chaos. The results obtained from this conjunction of Theory of Writing, Corpus Linguistics and Graphematics, all in the service of Historical Linguistics, indicate – based on the analysis of the 23 patterns applied to the official letters of Rio Grande do Norte – a graphical reality in Brazil from the 18th and 19th centuries, as well as the first half of the 20th century that can be characterized, in lines with what Ramírez Luengo (2012b, 2013) observes on Spanish-American colonial manuscripts, by the coexistence of different sets of socially accepted graphic solutions at specific moments, and, in spite of what is traditionally claimed, this reality produced not a state of confusion/ chaos, but a range of graphical preferences/ tendencies drawn along the 250 years of written language examined which point to a process of orthographic standardization, not yet completed, however, up until the 1950s, the final chronological limit of the corpus under study.