This thesis adopts the theoretical framework of contrastive textology (CT) developed by Hartmann (1980) for the analysis of the language of French and English expository texts from the domain of social statistics CT results from a combination of two linguistic orientations text linguistics and contrastive stylistics (CS). Hartmann uses the term parallel texts to describe (a) translated texts and (b) non-translated texts in two languages which were produced in circumstances so similar as to produce comparable linguistic features. Although translation techniques and text-linguistic norms can more readily be observed and compared in translated texts, non-translated texts have two advantages they allow us to observe text-linguistic features in the two languages prior to any translation activity, and they reflect no influences of a source text. The aims of the thesis were threefold firstly to provide a descriptive inventory of some of the most important linguistic features of social statistics which would benefit students and teachers of French, secondly to provide guidelines for the translation of social statistics texts from French into English, and finally to determine whether parallel non-translated texts can be used effectively to provide guidelines for translation. To these ends some of the most important lexical, grammatical, syntactic, semantic and textual features of the social statistics texts were analyzed and compared. As an intended aid to language teaching and learmng, comprehensive frequency listings of many semantic categories of words were drawn up In terms of translation guidelines, it was advised that translators take note of the following differences between the languages French social statistics (a) has far more umque words, (b) uses a considerably higher proportion of sentence connecters and (c) uses rather more demonstrative noun phrases than English social statistics. The translation guidelines were then tested on a published professional translation and were found to hold true. The thesis concludes that in spite of some drawbacks, parallel nontranslated texts can be used with some success in establishing guidelines for translation.