Normalization is the point of departure of Dominic Stewart's (2000) corpus study of conventionality in the context of teaching L2 translation. The aim of his study was to test the following explanatory hypothesis: using a large general-language monolingual corpus of English as an aid to translating into English as a foreign language leads to the reproduction of conventional patterns in L2 translations. Stewart's classroom-based research shows that the use of the British National Corpus (BNC) for translating tourist brochures from Italian into English as L2 encourages students to look for and produce natural-sounding collocations, by examining the frequency of occurrence and concordance lines of assumed target language equivalents of source language noun phrases. Two examples of fluent collocations retrieved from the BNC are grand tour of the city, as the equivalent of gran giro della città and road with panoramic views, as the equivalent of strada panoramica. A large, general-language corpus such as the BNC can therefore be a very useful resource for students translating into English as a foreign language since it can compensate for their lack of native-speaker knowledge of target language and culture. However, argues Stewart, the use of corpora in the translation classroom may well contribute to reinforcing the normalizing tendency displayed by translated texts and perhaps inhibit creativity. In order to test this hypothesis Stewart suggests carrying out a comparative examination of lexical and syntactic conventionality between translations produced using a reference general-language corpus, such as the BNC, and translations produced using traditional resources alone.

The Unique Item Hypothesis has been tested experimentally in the translation classroom by Pekka Kujam?ki (2004) with a view to raising student awareness of what translation entails, through the examination of their own translations. In the first phase of the experiment, thirty-six students were asked to back-translate into Finnish the German and English translations of a Finnish original text created ad hoc on the topic of driving in Finland, which included several language-specific items with no straightforward equivalents in either German or English. In the second phase of the experimental design the students' translations were compared with the students' use of original Finnish, as revealed by a cloze test designed to elicit unique items. The findings confirmed the Unique Items Hypothesis. In their translations, in fact, students tended to overlook unique items and opt for straightforward lexical or dictionary equivalents, even if TL-specific items were found to be part of their lexical repertoire, as revealed by the results of the cloze test.

Finally, within the research area of translation quality assessment (TQA), Federica Scarpa (2006) has assessed, in relation to specialized translations carried out by advanced translator trainees, the validity of simplification and explicitation as translation universals and possible indicators of translation quality. The aim of the study is twofold: to deepen our understanding of translational behaviour in a particular training context and to provide reliable measures of quality with a view to improving the effectiveness of translator training. Specialized English–Italian translations carried out by advanced translator trainees were first compared with the English source texts as regards overall length, number of sentences, average sentence length, standardized type/token ratio, and lexical density. The results of this initial analysis were then compared with the TQA grades awarded. In the last stage, the translated texts were compared with comparable originals drawn from the Italian reference corpus CORIS (Corpus dell' Italiano Scritto). Italian translations were generally found to have more running words as well as fewer and longer sentences than the English originals, type/token ratio was higher and lexical density lower. Compared to lower-scoring translations, higherscoring translations were found to have a lower number of running words, higher average sentence length, lower number of sentences, higher type/token ratio, and lower lexical density. Higher-scoring and lower-scoring translations deviated from comparable originals to the same extent on all measures. The validity of simplification and explicitation as S- and T-universals was largely confirmed by comparisons made with source texts and comparable originals. The analysis of the relation between universals and TQA grades showed that higher-scoring translations have a higher level of explicitness and a lower level of simplification compared with lower-scoring translations.

All three studies explicitly draw on translation universals as empirically established features of actual behaviour. They test experimentally and in different classroom environments a number of explicit and implicit hypotheses, adopting either a comparative or a causal research model. Stewart’s hypothesis is explicitly explanatory, while Kujam?ki's is explicitly descriptive. Scarpa adopts a causal model and her research design is more complex. First of all she puts forward a number of interpretive and descriptive hypotheses aimed at testing the validity of universals as attested in the literature. She then suggests an implicit two-tailed explanatory hypothesis, which can be phrased as: either a high or a low TQA score is caused by the presence of certain translation profile features, which are regarded as aspects of universals of translation. Finally, she tests this explanatory hypothesis by putting forward an implicit two-tailed predictive hypothesis of effect, which can be phrased as: a given set of translation profile features, which are viewed as aspects of translation universals, will result in either a high or a low TQA score.

責(zé)任編輯：admin