This project investigates to what extent register influences language use in Dutch translations in the same way as it influences non-translated Dutch. Building on the Dutch Parallel Corpus, we apply profile-based correspondence analysis to get a better insight in the patterns of variation.
Summary of the PhD thesis (by Isabelle Delaere; see publications)
This thesis sets out to investigate the long-standing assumption that translated language as such differs from non-translated language by means of operationalizing an umbrella hypothesis which predicts that translated language will conform to the norm more often than non-translated language. Moreover, so as to take the multidimensional character of translations into account we not only investigated the factor translation status, but also two additional, extralinguistic factors, viz. genre and source language.
These factors were analyzed with regard to two types of norm conformity: descriptive and prescriptive norm conformity. The difference between these two types of norms can be summarized as follows: descriptive norms describe how things are done whereas prescriptive norms prescribe how things should be done. This thesis focuses on descriptive norm-conforming behavior by examining to what degree translations conform to genre-determined norms which describe what is prevailing language usage in a given genre and what is not. We investigate prescriptive norm-conforming behavior on the other hand by verifying to what extent translations and non-translations conform to the linguistic norms issued by the Dutch Language Union which prescribe whether certain lexemes, expressions, etc. belong to the standard language or not.
So as to be able to visualize the norm-conforming behavior in various genres and source language varieties we applied profile-based correspondence analysis which is an exploratory, multivariate statistical technique. As lexical variation is based on the idea that language users can choose between two or more formal alternatives to express a given concept, a profile-based approach is very well suited to investigate these linguistic preferences as the analysis is based on the difference in use between the formal alternatives and not the concepts behind them. Moreover, the technique leads to conclusions which are valid on a general level due to the objectivity of the statistical method and the substantial number of variables which can be used in each analysis and which are analyzed together instead of separately.
In order to investigate norm-conforming behavior four sub-hypotheses were derived from the central norm conformity hypothesis, each of which was operationalized by means of a case study. More specifically, two case studies were carried out which focused on exploring descriptive norm conformity, and two case studies were carried out to investigated prescriptive norm conformity. Although the present study suffered from a number of limitations, the empirical findings from our corpus-based multivariate approach showed that translations do not always walk the line as they do not necessarily display more norm conform behavior than comparable non-translations and that the factors genre and source language and all underlying aspects affect the degree to which a given translation conforms to the norm.