Le présent article met en évidence deux fonctions distinctes du corpus : heuristique et validante. Chacun connaît la seconde : un corpus est utilisé afin de démontrer la plausibilité ou la véracité d'une hypothèse qui a une existence indépendante. A la différence de celle-ci, la première fonction montre, plutôt qu’elle ne démontre. Le linguiste ne cherche plus ici à convaincre ; il cherche à comprendre. Il part à la rencontre de l'inconnu avec un questionnement, mais sans solution, même hypothétique, pour le puzzle dont il a identifié quelques pièces. La vertu du corpus ici est consubstantielle à son existence : le classement. Qui dit corpus dit en effet principe de classement. Le corpus heuristique a ceci de particulier qu'il est éphémère : sitôt ses services rendus, il sera transformé par le linguiste en corpus de validation (qui aura vocation à convaincre le monde). Ce changement d'objectif induit alors une organisation et une présentation très différentes de celles qui ont permis la découverte du fonctionnement linguistique en question. En remontant cette opération, le présent article essaie d'isoler les propriétés invariables des corpus heuristiques. Le propos est illustré par un cas particulier tiré de mon expérience personnelle, i.e. les alternances de longueur vocalique dans les itératifs tchèques.
The heuristic corpus : a tool that shows the way, but does not demonstrate anything
This article proposes to recognise two different functions that a corpus can endorse: heuristic and validating. The latter is common: a corpus is used in order to demonstrate the plausibility or the validity of a hypothesis which has an independent existence. By contrast, the former shows the way, rather than demonstrating anything. The linguist does not try to convince anybody - he tries to understand. He ventures into unknown land with a question, but without solution, not even with a hypothesis, holding in hands only a few pieces of the puzzle that needs to be put together. It is the corpus itself which will provide new pieces, a hypothesis and, ideally, a solution. The corpus acts as a counsellor, makes appear the "reality" in a new light and restricts the pool of possible solutions. The property of the corpus which allows it to play this role is consubstantial with its existence: classification. There is no corpus without classification of the data at hand. And it is a peculiar classification that will shed light on the "reality" in a way so to produce the "heureka" situation on the linguist's side. As a consequence, there could not be any corpus, in linguistics or elsewhere, without a project of exploitation or without theoretical activity prior to its actual genesis. Any corpus, of the heuristic as much as of the validating kind, is an oriented, hence a theoretical object.
The heuristic corpus is also peculiar in that it is promised to oblivion: as soon as it has served, the linguist will transform it into a validating corpus (he wants to convince the world). This change of orientation supposes an organisation and a presentation that are quite different from those that have promoted the discovery in the first place. Therefore, the present article also aims at rescuing the heuristic corpus from oblivion: after the battle, its only existence are often some sheets of paper in a forgotten folder or some old file lost on a harddisk, and the linguist will have completely forgotten that there is any such file, and how it helped him to discover what now has become a reality. I attempt at isolating some invariable properties of heuristic corpora: the moment when they enter the scene (i.e. when the linguist is facing a dead-end), the situations where they may be helpful and those where they will not advance the research.
The overall purpose of the article is illustrated by a case study from my own experience, i.e. the alternations in vowel length that are found in Czech iteratives.