«[...] la palabra [...] es en sí misma metafórica y por ello
imprecisa [...] »¹.

The translation into words of the mental content the person writing wants to express is the subject of studies in the field of psycholinguistics, where this stage also specific to the interlingual translation process is called "text generation". Researchers dealing with it implicitly agree on the fact that it is a translation process - within the wider interlingual translation process - and to describe it spontaneously use the word "translate" and its derivatives:

Text-generation researchers investigate the way in which the formulation of a speech act is obtained starting from the formation of the material to be expressed through the intention to express it. In the case of interlingual translation, the text-generation process passes through one less stage: we can consider the content to be expressed already formed since, presumably, it is derived from the prototext reading as described in the second part of this course. We are, therefore, interested in all the stages of text generation following content formation.
  We wonder whether it makes any sense to so neatly limit the notion of translation. Didn''t we see that the total translation view greatly broadens the semantic field of the word "translation"? Does it make any sense even so to state that the content expressed in a metatext is nonetheless derived from the prototext? The most sensible answer is affirmative. "Total translation" does not mean at all "free writing" and, if transfers or semiotic changes originally outside the focus of translation theory are now its subjects, the point is that we have a prototext and a metatext, and the latter is an elaboration of the former.
  What exactly comprises this elaboration is not specified, but on the fact that it is an elaboration - not an invention from nothing, if this is possible - there are no doubts.
  Human mind processes language by taking it apart - unconsciously: it''s all too fast for a conscious control to be active - into "processing units", conceptual chunks that may correspond to nominal groups, propositions, but never single words³. Word-for-word elaboration can lead to nowhere, can block sentence formulation. In a 1986 work, quoted in Zock 1997, the attention is focused, as an example, on what happens with French sentences containing clitics:

In the analyzed examples, it would be impossible to go on formulating the sentence not knowing from start what person is the object of the action of giving⁴.
  The length of the chunks used by the single individual as a processing unit depend, beyond the complexity of the concepts to be expressed, on the technical competence of the person writing. The more expert and able the writer is, the better she is able to work on extended chunks. A translator, while writing the metatext, is no exception to this descriptive rule⁵: The more technical experience she has in translations, the greater the text chunks she applies at first translation microstrategies to.
  By translating a mental chunk⁶ into words, the person writing projects a conceptual structure (deep structure) onto a linguistic form (surface structure). Concepts are projected onto words, each of which has a grammatical category identity (its part of speech: noun, verb, adjective, adverb etc.), an immediate relational identity (subject, direct object, indirect object) and a larger relational identity (kind of sentence)⁷. These three degrees of complexity correspond to larger and larger chunks and require increasingly complex planning skills.
  The content and lexical form of the speech acts produced does not depend on the mental material but only on its base. Expressive "habits", "customs", cognitive experiences that tend to be repeated must be considered.

A first lexical draft of the mental content to be expressed is sometimes realized in this way, although looking for approximate matching between previous writing experiences and what needs to be expressed. This pattern matching produces a first approximate draft, comparable to what Freud calls "primary processing" referring to dream lexicalization. Such first stage implies, especially in the more expert and skilled writers, a second stage of reviewing and adjusting.

  

Bibliographical references

BATEMAN J. & ZOCK M. Natural Language Generation, in R. Mitkov, editor, Handbook of Computational Linguistics, Oxford University Press, 2001, ISBN

MARÍAS J. Negra espalda del tiempo, Punto de lectura, 2000 (original edition 1998), ISBN 84-663-0007-7.

ZOCK M. Holmes meets Montgomery: an unusual yet necessary encounter between a detective and a general, or, the need of analytical and strategic skills in outline planning, in VI Simposio Internacional de Comunicacion Social, Santiago de Cuba, 1999, p. 478-483.

ZOCK M. The power of words, in Message Planning, 16th International Conference on Computational Linguistics (COLING), Copenhagen, 1996, p. 990-5.

ZOCK M. Sentence generation by pattern matching: the problem of syntactic choice, in R. Mitkov & N. Nicolov editors, Recent Advances in Natural Language Processing. Series: Current Issues in Linguistic Theory, Amsterdam, Benjamins, 1997, ISBN p. 317-352.

---

¹ Marías 1998 (2000), p. 10.
² Added bold. Zock 1997, p. 317.
³ Zock 1997, p. 318.
⁴ Zock 1997, p. 318-319, note.
⁵ Some computer-assisted-translation systems, like Wordfast, do not use words as units; they use more extended chunks. Translators on a case-by-case basis can customize their measure.
⁶ Some authors speak of mental "images". Some prefer to use generic terms in order to avoid attributing inner language - characterized by many different semiotic codes - an exclusively visual dimension.
⁷ Zock 1997, p. 321-322.
⁸ Zock 1997, p. 323. Bold added.