OXFORD-HACHETTE FRENCH DICTIONARY. Oxford University PressHachette. pound;28.99.

These two major French dictionaries are so obviously rivals that one can almost hear them snarl at one another across the desk.

Not that they are often likely to find themselves on the same desktop, being heavy-duty reference works that most people would not wish to duplicate. They invite comparisons, they insist that you choose.

Inevitably, the choice is not easy. The Collins-Robert is attractively produced, with a distinctive typeface that makes for a lighter, less crowded page. There is a distinct effort by CR to speed up word-hunting, which includes the introduction of a “menu” at the start of the most complex entries (for example: for, do, faire, que) showing how the article is divided; but it will be of little use to anyone who cannot tell the difference between a transitive and intransitive verb or, in the French-English section, understand the terms verbesubstitut or verbe copule.

In both dictionaries, grammatical explanations are given in the language of the headword; and neither seems to be aware of how far one can go nowadays in the study of a modern language without being encumbered by the technicalities of traditional grammar.

In this respect, they do not differ very much from similar dictionaries of 50 or more years ago. The chief novelty, in both cases, is that here the teams of compilers each worked from a (different) electronic corpus containing huge numbers of examples of words in use: for more information, see the copious explanations in the prefaces to the dictionaries themselves. “Was it worth it?” the Oxford-Hachette asks, in what is clearly intended to be a rhetorical question - or question rhetorique as the French would say.

But would they? OH confidently gives this as the translation, but it appears in none of my monolingual dictionaries and I cannot remember seeing it used in French; the Collins-Robert gives question pour la forme or l’effet. The next question, which is far from rhetorical, is why, if each of these works has employed its electronic database “much as scientists analyse objective data to test their hypotheses” (CR), the results they have obtained are so often dissimilar?

In its advice on using the dictionary, OH leads the reader through seven sample translations of phrases in the two languages - but following exactly the same process with CR would produce a quite different outcome in two of the seven cases, and could easily do so in a further two. So, translating the sentence he treated her kindly would give il l’a traitee avec gentilesse if you choose OH, and avec bienveillance if you prefer CR; a “sophisticated” nightclub would be chic in OH and raffine in CR.

The Collins-Robert is especially pleased with its inclusion of up-to-date terms such as alcopop (not in OH); while the Oxford-Hachette boasts “an encyclopedic A-Z of contemporary France”; but these are trimmings: the reader can also turn elsewhere for maps, a list of irregular verbs, a course on essay-writing and samples of everyday correspondence.

What one wants from a dictionary is a simple, unambiguous answer to the question: “How do I translate...?”; and in the days when dictionaries were written by linguists using their own intuition plus whatever examples they had to hand, this is what you got (if the usage of native speakers then proved to be different, the native speakers had got it wrong). Such certainties no longer exist: the database and its corpora, far from conferring greater authority on these works, merely underline the shifting nature of language and the impossibility of translation.

Yet we need bilingual dictionaries; and each of these is a mine of information and a product of admirable scholarly labour. Since a decision is inevitable, I would have a slight preference for the Collins-Robert: it is cheaper and wears its erudition more lightly: typography and layout make it a little easier to use.

The Oxford-Hachette, on the other hand, is stronger on literary language, with agenerally more solid appearance, a greater number of references and better cross-referencing. CR has looke, OH has Oulipo; CR has “pantyhose” and “cyberpet”, OH has “sit-ups” and “sjambok”. But neither inspires total confidence in the user: their combined vocabulary is vast, but do you trust them to know how to use it?