In this chapter, we will describe the effects of the software keyboard layout representation. Our demonstration will rely on several characteristics of software keyboards: morphologic features (form, size, space between keys) and several spatial representations (QWERTY, AZERTY, alphabetic, etc.) designed by our software platform toolkit. Theoretical results obtained by means of several models (Mackenzie and Zhang 1999, Vella and Vigouroux 2009) based on Fitts's law (Fitts 1954) will be discussed. Then, we will report the experiment on the VITIPI word completion system, (Boissière et al. 2012), to demonstrate the impact of the corpora size and the linguistic register on the performance of the completion system through the Key Stroke Rate, (Boissière et al. 2006). This study demonstrates that a user specific linguistic representation is, very often, more accurate than a general representation. We will also conclude that the number of words ``out of vocabulary'' is also important. Finally, we will describe a neuro-psycholinguistic approach to categorize written error, (Boissière et al. 2007), on the following units: letter, grapheme and morpheme. This categorization has been elaborated on text corpora produced by people with different types of language disorders.