ABSTRACT
Essential to the method of van Leuven-Zwart is the priority given to the concept of relation. Any comparative description involves establish ing the relation between elements as well as attributing certain features to those elements. According to van Leuven-Zwart, a comparison based on the prior attribution of features is only a ' second degree comparison' , since it departs from a descriptive operation, and the relation ship between the elements is established afterwards. In a direct comparison, the order is reversed. According to this scheme, a texteme, for instance, would be a unit of description rather than a unit of comparison. At the microstructural level, a relation of complete conjunction between the transemes and archi transemes (in which case there is a relation of synonymy between transemes) is assumed as a starting point, and shifts occur when there are aspects of disjunction between transemes and the A TR. Van Leuven-Zwart distinguishes three main categories: modulation (where a source or target transeme shows one or more aspects of disjunction with the A TR; a relation of hyponymy between transemes), modification (where a source and target transeme show one or more aspects of disjunction with the A TR; a relation of contrast between transemes), and mutation (where there are no aspects of con junction, and therefore no A TR can be established; no relation between transemes ). The purpose of this method is to arrive at hypotheses about the interpretation and the strategy underlying the translation involved in the comparison. As a consequence, shifts that do riot reflect a translator's interpretation or strategy are not taken into account: only optional shifts and substantial shifts are consid ered. As far as the distinction between obligatory and optional shifts is concerned, van Leuven-Zwart expresses a reservation as to its applicability. In the first instance, the decision whether shifts are to be considered optional or obligatory is suspended. Not until the effects of the microstructural shifts on the macrostructural level have been established will it be possible to determine to what extent the shifts are due to other than purely linguistic factors. Initially , all substantial shifts are noted; that is, all the shifts that have some bearing on one of the substantial levels, namely the semantic, stylistic or prag matic level. As to syntactic shifts, only those
that affect these substantial levels are taken into account. Purely formal shifts are disregarded. On the distinction between formal and substan tial shifts, see also van den Broeck and Lefevere (1979).
